Overview

An important and probably highly desirable feature is that ANNOVAR can help identify subsets of variants based on comparison to other variant databases, for example, variants annotated in dbSNP or variants annotated in 1000 Genome Project. The exact variant, with same start and end positions, and with same observed alleles, will be identified.

These functionalities mentioned above can be performed using the --filter operation in ANNOVAR. The major difference between --filter and --regionanno above is that that --filter operation works on mutations (nucleotide changes), but --regionanno operation works on chromosome locations. For example, --region compare variants with things like chr1:1000-1000, but --filter compare variants with things like A->G change at the position chr1:1000-1000.

Summary of databases

Due to the increased number of databases that are available at ANNOVAR, some users are not sure where to start. Here we give a brief summary of some of the mostly commonly used databases.

For frequency of variants in whole-genome data:

  • 1000g2015aug: latest 1000 Genomes Project dataset with allele frequencies in six populations including ALL, AFR (African), AMR (Admixed American), EAS (East Asian), EUR (European), SAS (South Asian). These are whole-genome variants.
  • kaviar_20150923: latest Kaviar database with 170 million variants from 13K genomes and 64K exomes.
  • hrcr1: latest Haplotype Reference Consortium database with 40 million variants from 32K samples in haplotype reference consortium
  • cg69: allele frequency in 69 human subjects sequenced by Complete Genomics. useful to exclude platform specific variants.

For frequency of variants in whole-exome data:

  • exac03: latest Exome Aggregation Consortium dataste with allele frequencies in ALL, AFR (African), AMR (Admixed American), EAS (East Asian), FIN (Finnish), NFE (Non-finnish European), OTH (other), SAS (South Asian).
  • esp6500siv2: latest NHLBI-ESP project with 6500 exomes. Three separate key words are used for 3 population groupings: esp6500siv2_all, esp6500siv2_ea, esp6500siv2_aa.

For frequency of variants in isolated or less represented populations:

  • ajews: common alleles in ashkenazi jews
  • TMC-SNPDB: common alleles in Indian populations
  • ???: I strongly encourage ANNOVAR users to donate your own allelel frequency database on isolated/special populations to be shared to worldwide users

For functional prediction of variants in whole-genome data:

  • gerp++: functional prediction scores for 9 billion mutations based on selective constraints across human genome. You can optionally use gerp++gt2 instead since it includes only RS score greater than 2, which provides high sensitivity while still strongly enriching for truly constrained sites
  • cadd: Combined Annotation Dependent Depletion score for 9 billion mutations. It is basically constructed by a support vector machine trained to differentiate 14.7 million high-frequency human-derived alleles from 14.7 million simulated variants, using ~70 different features. For known indels, use caddindel.
  • cadd13: CADD version 1.3.
  • dann: functional prediction score generated by deep learning, using the identical set of training data as cadd but with much improved performance than cadd.
  • fathmm: a hidden markov model to predict the functional importance of both coding and non-coding variants (that is, two separate scores are provided) on 9 billion mutations.
  • eigen: a spectral approach integrating functional genomic annotations for coding and noncoding variants on 9 billion mutations, without labelled training data (that is, unsupervised approach)
  • gwava: genome-wide annotation of variants that supports prioritization of noncoding variants by integrating various genomic and epigenomic annotations on 9 billion mutations.

For functional prediction of variants in whole-exome data:

  • dbnsfp30a: this dataset already includes SIFT, PolyPhen2 HDIV, PolyPhen2 HVAR, LRT, MutationTaster, MutationAssessor, FATHMM, MetaSVM, MetaLR, VEST, CADD, GERP++, DANN, fitCons, PhyloP and SiPhy scores, but ONLY on coding variants

For functional prediction of slice variants:

  • dbscsnv11: dbscSNV version 1.1 for splice site prediction by AdaBoost and Random Forest, which score how likely that the variant may affect splicing
  • spidex: deep learning based prediction of splice variants. Unlike dbscsnv11, these variants could be far away from canonical splice sites

For disease-specific variants:

  • clinvar_20160302: ClinVar database with separate columns (CLINSIG CLNDBN CLNACC CLNDSDB CLNDSDBID) for each variant
  • cosmic70: the latest COSMIC database with somatic mutations from cancer and the frequency of occurence in each subtype of cancer. For more updated cosmic, see instructions below on how to make them.
  • icgc21: International Cancer Genome Consortium version 21 mutations.
  • nci60: NCI-60 human tumor cell line panel exome sequencing allele frequency data

For variant identifiers:

  • snp142: dbSNP version 142
  • avsnp142: an abbreviated version of dbSNP 142 with left-normalization by ANNOVAR developers.

1000 Genomes Project (2015 Aug) annotations

See general instructions below, but changing command line argument to 1000g2015aug.

This set is almost identical to the 2014oct version. However, an ANNOVAR user identified bugs in chrX frequency in 1000G 2014oct data, which I traced back to the distribution of the original data provided by 1000G. The 1000G team has fixed this bug on 8/18/2015, which is now relected in the 1000g2015aug dataset in ANNOVAR.

1000 Genomes Project (2014 Oct) annotations

See general instructions below, but changing command line argument to 1000g2014oct. Note that ANNOVAR does provide 1000g2014sep and 1000g2014aug, but they are obselete now!

The 2014Oct data set contains alternative allele frequency data in 1000 Genomes Project for autosomes and sex chromosomes (ALL, AFR (African), AMR (Admixed American), EAS (East Asian), EUR (European), SAS (South Asian)). The data is based on 201409 collection v5 (based on 201305 alignment). See below for detailed usage instructions on a different version of the dataset.

1000 Genomes Project (2012 April) annotations (obselete!)

This is based on phase 1 release v3 called from 20101123 alignment. The populations include ALL, AMR, AFR, ASN and EUR. In addition, I also performed liftOver and provide these data sets in hg18 coordinate, to faciliate researchers working on hg18 coordinates.

To download the database, use following:

[kaiwang@biocluster ~/]$ annotate_variation.pl -downdb 1000g2012apr humandb -buildver hg19
NOTICE: Web-based checking to see whether ANNOVAR new version is available ... Done
NOTICE: Downloading annotation database http://www.openbioinformatics.org/annovar/download/hg19_1000g2012apr.zip ... OK
NOTICE: Uncompressing downloaded files
NOTICE: Finished downloading annotation files for hg19 build version, with files saved at the 'humandb' directory

To annotate a data set called ex1.avinput by the database and generate output file with name starting from ex1, use following:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype 1000g2012apr_eur -buildver hg19 -out ex1 example/ex1.avinput humandb/
NOTICE: Variants matching filtering criteria are written to ex1.hg19_EUR.sites.2012_04_dropped, other variants are written to ex1.hg19_EUR.sites.2012_04_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 2766067 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_EUR.sites.2012_04.txt...Done


[kaiwang@biocluster ~/]$ cat ex1.hg19_EUR.sites.2012_04_dropped
1000g2012apr_eur 0.04 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
1000g2012apr_eur 0.87 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
1000g2012apr_eur 0.81 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
1000g2012apr_eur 0.06 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
1000g2012apr_eur 0.54 1 84875173 84875173 C T comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
1000g2012apr_eur 0.96 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
1000g2012apr_eur 0.05 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
1000g2012apr_eur 0.01 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
1000g2012apr_eur 0.01 16 50763778 50763778 - C comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2
1000g2012apr_eur 0.53 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

The command above annotate the ex1.hg19.avinput file against 1000 Genomes Project 2012 April release on European subjects. Known variants will be written to the *dropped file together with allele frequencies. The variants without matching database entries will be written to the *filtered file.

It is possible to apply a MAF threshold to the filtering procedure:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype 1000g2012apr_eur -buildver hg19 -out ex1 example/ex1.avinput humandb/ -maf 0.05
NOTICE: Variants matching filtering criteria are written to ex1.hg19_EUR.sites.2012_04_dropped, other variants are written to ex1.hg19_EUR.sites.2012_04_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 2766067 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_EUR.sites.2012_04.txt...Done


[kaiwang@biocluster ~/]$ cat ex1.hg19_EUR.sites.2012_04_dropped
1000g2012apr_eur 0.87 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
1000g2012apr_eur 0.81 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
1000g2012apr_eur 0.06 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
1000g2012apr_eur 0.54 1 84875173 84875173 C T comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
1000g2012apr_eur 0.05 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
1000g2012apr_eur 0.53 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

This means that only variants whose allele frequency is higher or equal to 0.05 are printed to the *dropped file.

You can also reverse this threshold:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype 1000g2012apr_eur -buildver hg19 -out ex1 example/ex1.avinput humandb/ -maf 0.05 -reverse
NOTICE: Variants matching filtering criteria are written to ex1.hg19_EUR.sites.2012_04_dropped, other variants are written to ex1.hg19_EUR.sites.2012_04_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 2766067 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_EUR.sites.2012_04.txt...Done


[kaiwang@biocluster ~/]$ cat ex1.hg19_EUR.sites.2012_04_dropped
1000g2012apr_eur 0.04 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
1000g2012apr_eur 0.96 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
1000g2012apr_eur 0.05 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
1000g2012apr_eur 0.01 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
1000g2012apr_eur 0.01 16 50763778 50763778 - C comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2

In this case, only rare variants that are observed in 1000G will be printed out to the *dropped file. Note that since -maf is used, a SNP with alternative allele frequency of 0.96 will also be printed, as its MAF is 0.04. To only check on alternative allele frequency, rather than MAF, you can use -score_threshold:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype 1000g2012apr_eur -buildver hg19 -out ex1 example/ex1.avinput humandb/ -score_threshold 0.05 -reverse
NOTICE: Variants matching filtering criteria are written to ex1.hg19_EUR.sites.2012_04_dropped, other variants are written to ex1.hg19_EUR.sites.2012_04_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 2766067 and the number of bins to be scanned is 15
NOTICE: Scanning filter database humandb/hg19_EUR.sites.2012_04.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_EUR.sites.2012_04_dropped
1000g2012apr_eur 0.04 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
1000g2012apr_eur 0.05 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
1000g2012apr_eur 0.01 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
1000g2012apr_eur 0.01 16 50763778 50763778 - C comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2

Similarly, to switch to other ethnicity groups, use 1000g2012apr_asn, 1000g2012apr_afr, 1000g2012apr_amr or 1000g2012apr_all as the database type. Let's try an example to see if the same sets of variants are observed in asians:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype 1000g2012apr_asn -buildver hg19 -out ex1 example/ex1.avinput humandb/
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ASN.sites.2012_04_dropped, other variants are written to ex1.hg19_ASN.sites.2012_04_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 2743052 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_ASN.sites.2012_04.txt...Done


[kaiwang@biocluster ~/]$ cat ex1.hg19_ASN.sites.2012_04_dropped
1000g2012apr_asn 0.13 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
1000g2012apr_asn 0.58 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
1000g2012apr_asn 0.83 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
1000g2012apr_asn 0.60 1 84875173 84875173 C T comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
1000g2012apr_asn 1.00 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
1000g2012apr_asn 0.33 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

You'll see that the R381Q in IL23R and R702W in NOD2 (both SNPs are pretty famous) are not found in Asians from the 1000 Genomes Project.

Technical notes: ANNOVAR has the ability to handle VCF file directly. Therefore, you can just directly interrogate 1000G data as a VCF database. For example, you can do this for 2011 May release of 1000G data:

[kaiwang@biocluster ~/]$ annotate_variation.pl inputfile ./ -vcfdbfile ALL.wgs.phase1.projectConsensus.snps.sites.vcf -filter -dbtype vcf

By default, allele frequency is used as second column in output. But you can use -infoasscore to use the INFO field in VCF file as printed score in output.

dbSNP annotations

ANNOVAR can identify the variant that are already reported in dbSNP and also identify the corresponding rs identifiers. To speed up dbSNP annotation, ANNOVAR provides pre-built indexed files in its database repository. Use -webfrom annovar in the command to download these files for use in annotation procedure. Only a few specific versions of dbSNP have been made available, so check download page for these specific versions. Finally, I wrote an article on dbSNP identifiers, and you are welcome to read and comment on it.

[kaiwang@biocluster ~/]$ annotate_variation.pl -downdb -buildver hg19 -webfrom annovar snp138 humandb
NOTICE: Web-based checking to see whether ANNOVAR new version is available ... Done
NOTICE: Downloading annotation database http://www.openbioinformatics.org/annovar/download/hg19_snp138.txt.gz ... OK
NOTICE: Downloading annotation database http://www.openbioinformatics.org/annovar/download/hg19_snp138.txt.idx.gz ... OK
NOTICE: Uncompressing downloaded files
NOTICE: Finished downloading annotation files for hg18 build version, with files saved at the 'humandb' directory

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -out ex1 -build hg19 -dbtype snp138 example/ex1.avinput humandb/
NOTICE: Variants matching filtering criteria are written to ex1.hg19_snp138_dropped, other variants are written to ex1.hg19_snp138_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 2858459 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_snp138.txt...Done


[kaiwang@biocluster ~/]$ cat ex1.hg19_snp138_dropped
snp138 rs35561142 1 11403596 11403596 - AT comments: rs35561142, a 2-bp insertion
snp138 rs149123833 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
snp138 rs1000050 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
snp138 rs1287637 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
snp138 rs11209026 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
snp138 rs6576700 1 84875173 84875173 C T comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
snp138 rs15842 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
snp138 rs80338939 13 20763686 20763686 G - comments: rs1801002 (del35G), a frameshift mutation in GJB2, associated with hearing loss
snp138 rs2066844 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
snp138 rs2066845 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
snp138 rs2066847 16 50763778 50763778 - C comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2
snp138 rs2241880 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

Two output files are generated. The ex1.hg19_snp138_filtered file contains SNPs not in dbSNP. The ex1.hg19_snp130_dropped file contains variants that are annotated in dbSNP, and print out their rs identifiers (as the second column)

Technical Notes: dbSNP 129 is generally regarded as the last "clean" dbSNP without "contamination" from 1000 Genomes Project and other large-scale next-generation sequencing projects.Per user request, I now make a dbSNP 129 in hg19 coordinate, so that users can actually use it to benchmark their variant calling algorithms, given that dbSNP 129 does not contain the "contaminations" from variants calls from next-generation sequencing.

Technical Notes: Since January 2011, per users' request, ANNOVAR now handles tri-allelic or quad-allelic SNPs. For example, rs12931472 can have four alleles (A, C, G, T) with wildtype as A, so any non-A mutation will be filtered by ANNOVAR, and rs12931472 will be printed out during filtering. In previous versions of ANNOVAR, only di-allelic SNPs are handled.

Technical Notes: These annotations may be assigned to "SNPs" in dbSNP: 'unknown','single','in-del','het','microsatellite','named','mixed','mnp','insertion','deletion'. ANNOVAR will only care about 'single', 'deletion', 'in-del', 'insertion' and ignore others. 'single' SNP accounts for the vast majority of dbSNP entries.

In 2012, sevearl additional "NonFlagged" dbSNP database are provided by me. Basically, these are dbSNP files subtracting Flagged dbSNP entries. Flagged SNPs include SNPs < 1% minor allele frequency (MAF) (or unknown), mapping only once to reference assembly, flagged in dbSnp as "clinically associated". Some users have reported that some SNPs are still flagged as "clinically associated" in the NonFlagged set; this is because these SNPs are not found in the Flagged set from UCSC, possibly because they are more recently associated with diseases so are not recorded in Flagged database yet.

The command line for downloading database and annotation is almost identical to regular dbSNP database. For example, to download the database:

[kaiwang@biocluster ~/]$ annotate_variation.pl -downdb -buildver hg19 -webfrom annovar snp135NonFlagged humandb
NOTICE: Web-based checking to see whether ANNOVAR new version is available ... Done
NOTICE: Downloading annotation database http://www.openbioinformatics.org/annovar/download/hg19_snp135NonFlagged.txt.gz ... OK
NOTICE: Downloading annotation database http://www.openbioinformatics.org/annovar/download/hg19_snp135NonFlagged.txt.idx.gz ... OK
NOTICE: Uncompressing downloaded files
NOTICE: Finished downloading annotation files for hg19 build version, with files saved at the 'humandb' directory

To give users an idea of the size of the various dbSNP databases prepared by me (first column represents number of variants). It is amazing that the number increased from 13.6 million in dbSNP129 to 63.3 million in dbSNP138 in just a few years!

13602827 hg19_snp129NonFlagged.txt
13610296 hg19_snp129.txt
18396965 hg19_snp130NonFlagged.txt
18404149 hg19_snp130.txt
25301548 hg19_snp131NonFlagged.txt
25312455 hg19_snp131.txt
32249106 hg19_snp132NonFlagged.txt
32267005 hg19_snp132.txt
53473344 hg19_snp135NonFlagged.txt
53502122 hg19_snp135.txt
55415160 hg19_snp137NonFlagged.txt
55449842 hg19_snp137.txt
63224151 hg19_snp138NonFlagged.txt
63289767 hg19_snp138.txt

The dbSNP entries does not include allele frequency measure, so users should exercise caution when using dbSNP as a filtering step to identify causal variants for Mendelian diseases, as some dbSNP entries may well be related to disease susceptibility.

Technical Notes: the dbSNP commonSNP track from UCSC is extremely incomplete and users really should not use it for annotation under any circumstance. Use the 1000g2014oct database or better the popfreq_all database provided in ANNOVAR for this purpose instead.

avSNP annotations

avSNP datasets are basically reformatted dbSNP datasets. Currently we made hg19_avsnp138 and hg19_avsnp142 available, but this list may expand in the future.

Unlike dbSNP, avSNP files have standard six-column "generic" format, and its indels have been left-normalized, and it contains only one variant per line. This makes the assignment of dbSNP identifiers much cleaner. For more detailed discussions, please read this article.

The use of avsnp datasets are straightforward with the avsnp138/avsnp142 as -dbtype:

annotate_variation.pl -downdb -webfrom annovar -buildver hg19 avsnp142 humandb/
annotate_variation.pl ex1.avinput humandb/ -filter -build hg19 -dbtype avsnp142

LJB* (dbNSFP) non-synonymous variants annotation

The LJB* databases (for historical reasons, it is named as ljb rather than dbNSFP in ANNOVAR) include SIFT scores, PolyPhen2 HDIV scores, PolyPhen2 HVAR scores, LRT scores, MutationTaster scores, MutationAssessor score, FATHMM scores, GERP++ scores, PhyloP scores and SiPhy scores. As of October 2015, the lastest ljb database is dbnsfp30a. Previously, this dataset is referred to as ljb26, ljb23, ljb2, ljb, which caused confusions among many users. Starting from version 3.0a, we will adopt the dbnsfp keyword for future updates. Additionally, users should start to use table_annovar.pl to calculate scores for non-synonymous variants, since we no longer provide individual scores for individual algorithms.

The command below takes an input file and generates 20 different scores and predictions for all the non-synonymous variants in the file. You can open the output file by Excel or other spreadsheet programs to examine.

annotate_variation.pl -downdb -webfrom annovar -buildver hg19 dbnsfp30a humandb/
table_annovar.pl ex1.avinput humandb/ -protocol dbnsfp30a -operation f -build hg19 -nastring .

The output include various scores in the following columns: SIFT_score SIFT_pred Polyphen2_HDIV_score Polyphen2_HDIV_pred Polyphen2_HVAR_score Polyphen2_HVAR_pred LRT_score LRT_pred MutationTaster_score MutationTaster_pred MutationAssessor_score MutationAssessor_pred FATHMM_score FATHMM_pred PROVEAN_score PROVEAN_pred VEST3_score CADD_raw CADD_phred DANN_score fathmm-MKL_coding_score fathmm-MKL_coding_pred MetaSVM_score MetaSVM_pred MetaLR_score MetaLR_pred integrated_fitCons_score integrated_confidence_value GERP++_RS phyloP7way_vertebrate phyloP20way_mammalian phastCons7way_vertebrate phastCons20way_mammalian SiPhy_29way_logOdds.

If a variant does not have a score (for example, for intronic variant), the corresponding position will be denoated by a period (which is specified by the -nastring argument above.

As mentioned above, for historical reasons, the dbNSFP database is referred to as LJB database, and we used to provide separate files for each individual scores. The description below refers to ljb23. They are helpful for users who only want to infer individual scores for individual prediction method.

The keyword used for downloading these data include: ljb23_sift, ljb23_pp2hdiv, ljb23_pp2hvar, ljb23_lrt, ljb23_mt, ljb23_ma, ljb23_fathmm, ljb23_metasvm, ljb23_metalr, ljb23_gerp++, ljb23_phylop, ljb23_siphy, ljb23_all. The ljb23_all includes ALL scores, and it is very useful in table_annovar.pl.

Detailed information for all the LJB23 databases are given below:

Score (dbtype) # variants in LJB23 build hg19 Categorical Prediction
SIFT (sift) 77593284 D: Deleterious (sift<=0.05); T: tolerated (sift>0.05)
PolyPhen 2 HDIV (pp2_hdiv) 72533732 D: Probably damaging (>=0.957), P: possibly damaging (0.453<=pp2_hdiv<=0.956); B: benign (pp2_hdiv<=0.452)
PolyPhen 2 HVar (pp2_hvar) 72533732 D: Probably damaging (>=0.909), P: possibly damaging (0.447<=pp2_hdiv<=0.909); B: benign (pp2_hdiv<=0.446)
LRT (lrt) 68069321 D: Deleterious; N: Neutral; U: Unknown
MutationTaster (mt) 88473874 A" ("disease_causing_automatic"); "D" ("disease_causing"); "N" ("polymorphism"); "P" ("polymorphism_automatic"
MutationAssessor (ma) 74631375 H: high; M: medium; L: low; N: neutral. H/M means functional and L/N means non-functional
FATHMM (fathmm) 70274896 D: Deleterious; T: Tolerated
MetaSVM (metasvm) 82098217 D: Deleterious; T: Tolerated
MetaLR (metalr) 82098217 D: Deleterious; T: Tolerated
GERP++ (gerp++) 89076718 higher scores are more deleterious
PhyloP (phylop) 89553090 higher scores are more deleterious
SiPhy (siphy) 88269630 higher scores are more deleterious

Some examples were given below:

- SIFT annotation

EXTREMELY IMPORTANT!!!!!! in first version of dbNSFP (ljb_sift), the scores were calculated as 1-SIFT. In the updated version 2 (ljb2_sift), the scores were now the SIFT score itself. In version 2.3 and later, the default scores are also the SIFT score itself, but you can optionally print out the "converted score" which is 1-SIFT, and the categorical prediction. This mean a variant with score<0.05 is predicted as deleterious.

In the example below, two missense variants were predicted as deleterious based on SIFT scores (<0.05). The command should take merely a few seconds.

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_sift -buildver hg19 -out ex1 example/ex1.avinput humandb/
NOTICE: the --dbtype ljb23_sift is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_sift_dropped, other variants are written to ex1.hg19_ljb23_sift_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 187938 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_sift.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_sift_dropped
ljb23_sift 0.1 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_sift 0.01 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_sift 0.02 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_sift 0.57 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

To show the categorical predictions in the output file, add the -otherinfo argument:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_sift -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_sift is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_sift_dropped, other variants are written to ex1.hg19_ljb23_sift_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 187938 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_sift.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_sift_dropped
ljb23_sift 0.1,0.90,T 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_sift 0.01,0.99,D 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_sift 0.02,0.98,D 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_sift 0.57,0.43,T 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

The three comma-separated numbers are raw score, converted score and categorical prediction, respectively. It is clear that the second and third variants are predicted as deleterious.

- PolyPhen 2 annotation

There are two databases for PolyPhen2: HVAR and HDIV. They are explained below:

  • ljb2_pp2hvar should be used for diagnostics of Mendelian diseases, which requires distinguishing mutations with drastic effects from all the remaining human variation, including abundant mildly deleterious alleles.The authors recommend calling "probably damaging" if the score is between 0.909 and 1, and "possibly damaging" if the score is between 0.447 and 0.908, and "benign" is the score is between 0 and 0.446.

  • ljb2_pp2hdiv should be used when evaluating rare alleles at loci potentially involved in complex phenotypes, dense mapping of regions identified by genome-wide association studies, and analysis of natural selection from sequence data. The authors recommend calling "probably damaging" if the score is between 0.957 and 1, and "possibly damaging" if the score is between 0.453 and 0.956, and "benign" is the score is between 0 and 0.452.

An example is given below:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_pp2hvar -buildver hg19 -out ex1 example/ex1.avinput humandb/
NOTICE: the --dbtype ljb23_pp2hvar is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_pp2hvar_dropped, other variants are written to ex1.hg19_ljb23_pp2hvar_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 184437 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_pp2hvar.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_pp2hvar_dropped
ljb23_pp2hvar 0.999 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_pp2hvar 0.901 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_pp2hvar 0.986 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_pp2hvar 0.005 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

As you can see, two of the four missense variants were predicted to have scores higher than 0.909, one has score higher than 0.447.

If you want to have the "probably damaging", "possibly damaging" and "benign" calls, you can add the -otherinfo argument:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_pp2hvar -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_pp2hvar is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_pp2hvar_dropped, other variants are written to ex1.hg19_ljb23_pp2hvar_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 184437 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_pp2hvar.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_pp2hvar_dropped
ljb23_pp2hvar 0.999,D 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_pp2hvar 0.901,P 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_pp2hvar 0.986,D 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_pp2hvar 0.005,B 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

In the output, the scores and predictions are separated by comma. There are three possible predictions: "D" ("porobably damaging"), "P" ("possibly damaging") and "B" ("benign").

Technical comments: The original LJB23 database provided by original authors may contain multiple annotations for each mutation. For example, the chr1:67705958G>A mutation has two separate records, one showing E>K amino acid substitution, and the other showing R>Q amino acid substitution. The former does not have any functional scores associated, so in my compilation, only the latter will enter into the database. Next, in the LJB2 database, even for R>Q substution, it can have multiple scores such as "1.0;1.0;1.0;1.0;0.993;1.0;1.0;1.0;0.999" and multiple predictions such as "D;D;D;D;D;D;D;D;D", probably due to multiple transcriptional isoforms. In this case, only the largest score (1 is the largest score among multiple isoforms), as well as its associated D/P/B annotation, will be used in ANNOVAR database.

- MutationTaster annotation

This is very similar to SIFT and PolyPhen. One example is shown below.

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_mt -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_mt is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_mt_dropped, other variants are written to ex1.hg19_ljb23_mt_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 193192 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_mt.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_mt_dropped
ljb23_mt 0.559,0.559,D 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_mt 1.000,0.000,N 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_mt 1.000,1.000,D 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_mt 0.001,0.999,P 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

In the output, the raw scores, converted scores and predictions are separated by comma. There are four possible predictions: "A" ("disease_causing_automatic"), "D" ("disease_causing"), "N" ("polymorphism") or "P" ("polymorphism_automatic").

- MutationAssessor annotation

Again very similar to SIFT and Polyphen. one example is given below:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_ma -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_ma is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_ma_dropped, other variants are written to ex1.hg19_ljb23_ma_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 179275 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_ma.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_ma_dropped
ljb23_ma 1.935,0.649,M 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_ma 2.32,0.683,M 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_ma 1.79,0.637,L 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_ma -0.255,0.459,N 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

Due to the use of -otherinfo argument, in the output, the scores and predictions are separated by comma. There are two possible predictions: predicted functional (H, M), predicted non-functional (L, N). Note that in ljb2, these are denoted as high/medium/low/neutral.

- LRT annotation

Again very similar to SIFT and Polyphen. one example is given below:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_lrt -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_lrt is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_lrt_dropped, other variants are written to ex1.hg19_ljb23_lrt_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 168542 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_lrt.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_lrt_dropped
ljb23_lrt 0.000,1.000,D 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_lrt 0.993,0.503,N 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_lrt 0.000,1.000,D 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_lrt 0.070,0.035,N 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

Due to the use of -otherinfo argument, in the output, the scores and predictions are separated by comma. There are three possible predictions: D(eleterious), N(eutral) or U(nknown).

- FATHMM annotation

Again very similar to SIFT and Polyphen. one example is given below:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_fathmm -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_fathmm is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_fathmm_dropped, other variants are written to ex1.hg19_ljb23_fathmm_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 178425 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_fathmm.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_fathmm_dropped
ljb23_fathmm 0.31,0.386,T 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_fathmm -0.62,0.421,T 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_fathmm 0.57,0.376,T 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_fathmm 0.81,0.367,T 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

If a score is smaller than -1.5 the corresponding NS is predicted as "D(AMAGING)"; otherwise it is predicted as "T(OLERATED)". All the four missense variants were predicted as tolerated by FATHMM. The method is less well known compared to SIFT and PolyPhen, but in our experience it works really well and better than SIFT/PolyPhen.

- MetaSVM annotation

MetaSVM is developed by Coco Dong at my lab in collaboration with Dr. Xiaoming Liu. It is composed of two steps: (1) perform imputation for whole-exome variants and fill out missing scores for SIFT, PolyPhen, MutationAssessor and so on. (2) Normalize all scores to 0-1 range (3) use a radial SVM model to train prediction model using all available scores and some population genetics parameters, and then apply the model on whole-exome variants. We used very high quality training data set for training the model, hence improving the performance of other algorithms.

It has clear advantage over other competing approaches such as SIFT/PolyPhen/CADD/Condel: (1) better performance (2) less missing values than SIFT/PolyPhen/Condel.

In a benchmarking study using manually compiled data set (basically variants published in Nature Genetics in 2011 as true positive), including 244 variants with half being disease causal, our method achieves best performance, compared to all other methods in ljb2 database (SIFT/PolyPhen/MutationTaster/MutationAssessor/LRT/FATHMM), as well as Condel/CADD/KGGSeq/MutPred/SNPGO/SNAP/PhDSNP/PANTHER/PONP. In another benchmarking study using the VariBench data, but excluding cancer mutations and then excluding all mutations reported in UniProt, totalling ~20,000 variants with 1/3 being disease causal, our method also achieves best performance compared to all other methods in ljb2. For more details, please refer to this paper.

An example to use the MetaSVM method is given below:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_metasvm -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_metasvm is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_metasvm_dropped, other variants are written to ex1.hg19_ljb23_metasvm_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 194687 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_metasvm.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_metasvm_dropped
ljb23_metasvm -0.599,0.351,T 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_metasvm -0.855,0.287,T 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_metasvm -0.696,0.327,T 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_metasvm -1.007,0.249,T 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

Similarly, the comma-delimited fields are raw score, converted score (0-1 range, higher score denoting more deleterious variants) and predictions. All the missense variants were predicted as tolerated. Note that the model building used variants that are known to cause Mendelian diseases so the method is specifically designed to work for Mendelian diseases but not complex diseases. These variants all lead to increased susceptibility to NOD2 with odds ratio<2, so they are not predicted as deleterious by MetaSVM.

- MetaLR annotation

Very similar to MetaSVM with similar performance. However, the model is far more interpretable.

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_metalr -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_metalr is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_metalr_dropped, other variants are written to ex1.hg19_ljb23_metalr_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 194687 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_metalr.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_metalr_dropped
ljb23_metalr 0.275,T 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_metalr 0.138,T 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_metalr 0.138,T 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_metalr 0.000,T 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

Only two comma-delimited fields are shown here, indicating raw score (0-1 range) and binary prediction. All variants were predicted as tolerated.

- GERP++ annotation

The ljb23_gerp++ contains only annotation for coding variants!!! If you want to annotate non-coding variants with GERP++ scores as well, read the section below.

Generally the higher the score, the more conserved the site.

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_gerp++ -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_gerp++ is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_gerp++_dropped, other variants are written to ex1.hg19_ljb23_gerp++_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 194631 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_gerp++.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_gerp++_dropped
ljb23_gerp++ 4.93 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
ljb23_gerp++ 5.19 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_gerp++ 3.66 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_gerp++ 5.91 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_gerp++ -11.4 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

The first three missense variants were predicted to be highly conserved and hence more likely to be functionally important.

- PhyloP and SiPhy annotation

Both are similar to GERP++ and these three can be considered as competitors of each other (just like SIFT and PolyPhen are competitors of each other).

PhyloP score is based on multiple alignments of 46 genomes. Similarly, SiPhy score is based on 29 mammals genomes. The larger the score, the more conserved the site.

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_phylop -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype ljb23_phylop is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_phylop_dropped, other variants are written to ex1.hg19_ljb23_phylop_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 195369 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_phylop.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_phylop_dropped
ljb23_phylop 0.745 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
ljb23_phylop 2.865 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_phylop 1.421 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_phylop 2.813 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_phylop -1.999 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype ljb23_siphy -buildver hg19 -out ex1 example/ex1.avinput humandb/ -otherinfo 
NOTICE: the --dbtype ljb23_siphy is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_ljb23_siphy_dropped, other variants are written to ex1.hg19_ljb23_siphy_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 192540 and the number of bins to be scanned is 7
NOTICE: Scanning filter database humandb/hg19_ljb23_siphy.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_ljb23_siphy_dropped
ljb23_siphy 6.573 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
ljb23_siphy 14.412 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
ljb23_siphy 6.914 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
ljb23_siphy 15.796 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
ljb23_siphy 12.660 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

As you can see, SiPhy differs from GERP++ and PhyloP, and predicted that the last variant is also highly conserved. Therefore, even conceptually similar tools may sometimes generate drastically different results. It is a good idea to use several techniques when doing funcitonal annotations of missense variants.

Notes: it is important for users to note that (1) these types of "conservation scores" only considers conservation level at the current base, and they do not care about the actual nucleotide identity so synonymous and non-synonymous variants at the same site will be scored as the same (2) these scores are not designed specifically for finding causal variants for Mendelian diseases, but for finding functionally important sites, so variants that confer increased susceptibility may be scored well.

- ljb23_all annotation

You can also just print out all scores for given variants using ljb23, by adding -otherinfo argument to the -filter annotation.

The columns are LJB23_SIFT_score LJB23_SIFT_score_converted LJB23_SIFT_pred LJB23_Polyphen2_HDIV_score LJB23_Polyphen2_HDIV_pred LJB23_Polyphen2_HVAR_score LJB23_Polyphen2_HVAR_pred LJB23_LRT_score LJB23_LRT_score_converted LJB23_LRT_pred LJB23_MutationTaster_score LJB23_MutationTaster_score_converted LJB23_MutationTaster_pred LJB23_MutationAssessor_score LJB23_MutationAssessor_score_converted LJB23_MutationAssessor_pred LJB23_FATHMM_score LJB23_FATHMM_score_converted LJB23_FATHMM_pred LJB23_RadialSVM_score LJB23_RadialSVM_score_converted LJB23_RadialSVM_pred LJB23_LR_score LJB23_LR_pred LJB23_GERP++ LJB23_PhyloP LJB23_SiPhy.

When using table_annovar.pl, it is best to use ljb23_all as the keyword, so that all these columns will be printed out automatically in the output file.

ESP (exome sequencing project) annotations

The ESP is a NHLBI funded exome sequencing project aiming to identify genetic variants in exonic regions from over 6000 individuals, including healthy ones as well as subjects with different diseases. The variant call data set is constantly being updated. ANNOVAR provide the allele frequency data so that users can evaluate if their variants have been previously observed in large population cohorts. As the size of the database is more than 1000 Genomes Project and the fold coverage is far higher, this data set will be particularly useful for users with exome sequencing data sets. As of October 2012, esp5400 and esp6500 are available, respresnting summary statistics from 5400 exoms and 6500 exomes, respectively. As of March 2015, the most recent version of ESP is esp6500siv2, so whenever possible, users should use this database for annotation. The instructions below was compiled for esp6500si.

The following commands download ESP6500si database for all ethnicity groups, then scan the ex1.avinput file against the database to find SNPs observed in ESP6500si:

[kaiwang@biocluster ~/]$ annotate_variation.pl -downdb -webfrom annovar -build hg19 esp6500si_all humandb/


[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype esp6500si_all -build hg19 -out ex1 example/ex1.avinput humandb/ 
NOTICE: the --dbtype esp6500si_all is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_esp6500si_all_dropped, other variants are written to ex1.hg19_esp6500si_all_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 191802 and the number of bins to be scanned is 8
NOTICE: Scanning filter database humandb/hg19_esp6500si_all.txt...Done


[kaiwang@biocluster ~/]$ cat ex1.hg19_esp6500si_all_dropped
esp6500si_all 0.027945 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
esp6500si_all 0.843133 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
esp6500si_all 0.046909 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
esp6500si_all 0.885787 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
esp6500si_all 0.007428 13 20763686 20763686 G - comments: rs1801002 (del35G), a frameshift mutation in GJB2, associated with hearing loss
esp6500si_all 0.031558 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
esp6500si_all 0.010157 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
esp6500si_all 0.016214 16 50763778 50763778 - C comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2
esp6500si_all 0.456251 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

To change to EA (European American) or AA (African American), just change the database name:

[kaiwang@biocluster ~/]$ annotate_variation.pl -downdb -webfrom annovar esp6500si_ea humandb/


[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype esp6500si_ea -build hg19 -out ex1 example/ex1.avinput humandb/ 
NOTICE: the --dbtype esp6500si_ea is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_esp6500si_ea_dropped, other variants are written to ex1.hg19_esp6500si_ea_filtered
NOTICE: Processing next batch with 12 unique variants in 12 input lines
NOTICE: Database index loaded. Total number of bins is 186024 and the number of bins to be scanned is 8
NOTICE: Scanning filter database humandb/hg19_esp6500si_ea.txt...Done


[kaiwang@biocluster ~/]$ cat ex1.hg19_esp6500si_ea_dropped
esp6500si_ea 0.029095 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
esp6500si_ea 0.817231 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
esp6500si_ea 0.063736 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
esp6500si_ea 0.956385 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
esp6500si_ea 0.010783 13 20763686 20763686 G - comments: rs1801002 (del35G), a frameshift mutation in GJB2, associated with hearing loss
esp6500si_ea 0.043488 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
esp6500si_ea 0.014535 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
esp6500si_ea 0.022171 16 50763778 50763778 - C comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2
esp6500si_ea 0.523953 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

The same six variants are found in esp6500si_all and esp6500si_ea database, but with different allele frequencies.

It is possible to use an allele frequency threshold. Unfortunately, the -maf_threshold argument is reserved for 1000G data set only. The -score_threshold argument can be used for all other databases. This is due to historical reasons.

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype esp6500si_ea -build hg19 -out ex1 example/ex1.avinput humandb/ -score_threshold 0.05
NOTICE: the --dbtype esp6500si_ea is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_esp6500si_ea_dropped, other variants are written to ex1.hg19_esp6500si_ea_filtered
NOTICE: Processing next batch with 12 unique variants in 12 input lines
NOTICE: Database index loaded. Total number of bins is 186024 and the number of bins to be scanned is 8
NOTICE: Scanning filter database humandb/hg19_esp6500si_ea.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_esp6500si_ea_dropped
esp6500si_ea 0.817231 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
esp6500si_ea 0.063736 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
esp6500si_ea 0.956385 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
esp6500si_ea 0.523953 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

Compared to the results above, the two SNPs with MAF<0.05 are no longer in the *dropped file. It is also possible to apply a -reverse argument:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype esp6500si_ea -build hg19 -out ex1 example/ex1.avinput humandb/ -score_threshold 0.05 -reverse
NOTICE: the --dbtype esp6500si_ea is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_esp6500si_ea_dropped, other variants are written to ex1.hg19_esp6500si_ea_filtered
NOTICE: Processing next batch with 12 unique variants in 12 input lines
NOTICE: Database index loaded. Total number of bins is 186024 and the number of bins to be scanned is 8
NOTICE: Scanning filter database humandb/hg19_esp6500si_ea.txt...Done


[kaiwang@biocluster ~/]$ cat ex1.hg19_esp6500si_ea_dropped
esp6500si_ea 0.029095 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
esp6500si_ea 0.010783 13 20763686 20763686 G - comments: rs1801002 (del35G), a frameshift mutation in GJB2, associated with hearing loss
esp6500si_ea 0.043488 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
esp6500si_ea 0.014535 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
esp6500si_ea 0.022171 16 50763778 50763778 - C comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2

Now, only rare SNPs that are observed in esp6500si_ea are printed to the *dropped file.

ExAC annotations

The Exome Aggregation Consortium (ExAC) is a coalition of investigators seeking to aggregate and harmonize exome sequencing data from a wide variety of large-scale sequencing projects. The allele frequency information for the ExAC can be downloaded using exac02, exac03 keywords in ANNOVAR. Currently supported population groups include ALL, AFR (African), AMR (Admixed American), EAS (East Asian), FIN (Finnish), NFE (Non-finnish European), OTH (other), SAS (South Asian)).

Technical Notes: unlike 1000G or esp6500 where allele frequency for each ethnicity group is contained within a separate file, the exac dataset in ANNOVAR include allele frequency information for all populations within a single file.

To download ExAC03 and examine allele frequeny for all variants within a file:

[kaiwang@biocluster ~/]$ annotate_variation.pl -downdb -webfrom annovar -build hg19 exac03 humandb/

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -build hg19 -dbtype exac03 example/ex1.avinput humandb/
NOTICE: the --dbtype exac03 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to example/ex1.avinput.hg19_exac03_dropped, other variants are written to example/ex1.avinput.hg19_exac03_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 749886 and the number of bins to be scanned is 10
NOTICE: Scanning filter database humandb/hg19_exac03.txt...Done

[kaiwang@biocluster ~/]$ cat example/ex1.avinput.hg19_exac03_dropped
exac03  0.054   1       1404001 1404001 G       T       comments: rs149123833, a SNP in 3' UTR of ATAD3C
exac03  0.825   1       5935162 5935162 A       T       comments: rs1287637, a splice site variant in NPHP4
exac03  0.041   1       67705958        67705958        G       A       comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
exac03  0.941   1       948921  948921  T       C       comments: rs15842, a SNP in 5' UTR of ISG15
exac03  0.006038        13      20763686        20763686        G       -       comments: rs1801002 (del35G), a frameshift mutation in GJB2, associated with hearing loss
exac03  0.023   16      50745926        50745926        C       T       comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
exac03  0.009917        16      50756540        50756540        G       C       comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
exac03  0.013   16      50763778        50763778        -       C       comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2
exac03  0.457   2       234183368       234183368       A       G       comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

The second column in the output file contains allele frequency for all samples (mixed ethnicity groups) within ExAC03.

To obtain frequency for each specific group, add -otherinfo argument:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -build hg19 -dbtype exac03 example/ex1.avinput humandb/ -otherinfo
NOTICE: the --dbtype exac03 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to example/ex1.avinput.hg19_exac03_dropped, other variants are written to example/ex1.avinput.hg19_exac03_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 749886 and the number of bins to be scanned is 10
NOTICE: Scanning filter database humandb/hg19_exac03.txt...Done

[kaiwang@biocluster ~/]$ cat example/ex1.avinput.hg19_exac03_dropped
exac03  0.054,0.0318,0.1525,0.1359,0.0874,0.0344,0.0783,0.0844  1       1404001 1404001 G       T       comments: rs149123833, a SNP in 3' UTR of ATAD3C
exac03  0.825,0.8949,0.7944,0.8179,0.9195,0.8304,0.8247,0.8305  1       5935162 5935162 A       T       comments: rs1287637, a splice site variant in NPHP4
exac03  0.041,0.0124,0.0321,0.0005,0.0555,0.0594,0.0541,0.0151  1       67705958        67705958        G       A       comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
exac03  0.941,0.7386,0.9632,0.9998,0.9681,0.9543,0.9410,0.9617  1       948921  948921  T       C       comments: rs15842, a SNP in 5' UTR of ISG15
exac03  0.006038,0.0013,0.0041,0,0.0103,0.0088,0.0088,0.0007    13      20763686        20763686        G       -       comments: rs1801002 (del35G), a frameshift mutation in GJB2, associated with hearing loss
exac03  0.023,0.0076,0.0177,0,0.0200,0.0349,0.0191,0.0004       16      50745926        50745926        C       T       comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
exac03  0.009917,0.0013,0.0071,0,0.0026,0.0141,0.0154,0.0083    16      50756540        50756540        G       C       comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
exac03  0.013,0.0027,0.0068,0.0005,0.0148,0.0202,0.0055,0.0015  16      50763778        50763778        -       C       comments: rs2066847 (c.3016_3017insC), a frameshift SNP in NOD2
exac03  0.457,0.3272,0.2130,0.3330,0.4387,0.5225,0.4989,0.5207  2       234183368       234183368       A       G       comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

The columns represent ExAC_Freq, ExAC_AFR, ExAC_AMR, ExAC_EAS, ExAC_FIN, ExAC_NFE, ExAC_OTH, ExAC_SAS, respectively.

Note that exac03 keyword can be used in table_annovar.pl directly to print out allele frequency for all ethnicity groups.

Technical Notes: Some users reported that the allele frequency measure differs slightly from the ExAC server itself, for some rare variants. This is because ExAC server displays "adjusted allele counts" and frequencies, yet ExAC03 within ANNOVAR displays the "AF" record in VCF file provided by ExAC. To address a growing number of user concerns, on 20150420, we finally decided to issue a new ExAC03 with ExAC_ALL header that shows the adjusted allele frequency for each variant to match the ExAC web site exactly.

Technical Notes: ExAC03 in ANNOVAR is one of the several data sets that have been left-normalized. See here for more details.

GERP++ annotations

GERP identifies constrained elements in multiple alignments by quantifying substitution deficits (see http://mendel.stanford.edu/SidowLab/downloads/gerp/ for details). I made annotation databases for all mutations with GERP++>2 in human genome, as this threshold is typically regarded as evolutioanrily conserved and potentially functional. Anything less than 2 is not informative, which helps reduce file size substantially.

The database (in hg19 or hg18 coordinate) needs to be downloaded first. Note that this is actually a large file (around 36GB) so it may take a while to download and decompress.

[kaiwang@biocluster ~/]$ annotate_variation.pl -downdb -buildver hg19 -webfrom annovar gerp++gt2 humandb/
NOTICE: Web-based checking to see whether ANNOVAR new version is available ... Done
NOTICE: Downloading annotation database http://www.openbioinformatics.org/annovar/download/hg19_gerp++gt2.txt.gz ... OK
NOTICE: Downloading annotation database http://www.openbioinformatics.org/annovar/download/hg19_gerp++gt2.txt.idx.gz ... OK
NOTICE: Uncompressing downloaded files
NOTICE: Finished downloading annotation files for hg19 build version, with files saved at the 'humandb' directory

Next to annotate an input variant file:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype gerp++gt2 -out ex1 -build hg19 example/ex1.avinput humandb/
NOTICE: the --dbtype gerp++gt2 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_gerp++gt2_dropped, other variants are written to ex1.hg19_gerp++gt2_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 1978403 and the number of bins to be scanned is 11
NOTICE: Scanning filter database humandb/hg19_gerp++gt2.txt...Done

Despite the large database size, it should take merely seconds to finish the annotation due to the use of indexing features. The conserved variants will be printed to the *dropped file (with the GERP++ scores), yet the other variants will be in the *filtered file.

[kaiwang@biocluster ~/]$ cat ex1.hg19_gerp++gt2_dropped
gerp++gt2 2.08 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
gerp++gt2 4.93 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
gerp++gt2 5.19 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
gerp++gt2 3.66 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
gerp++gt2 5.91 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2

One thing that is very important to note is that, since this is a filter operation, the program only looks to exact matches of SNPs. So if your input contains indels, it will not be in *dropped file.

In general, analyzing GERP++ scores for exome sequencing data should still be pretty fast, but it will be considerably slower for whole-genome seuqencing data due to the way ANNOVAR indexing is implemented.

For example, analyzing a file with 30726 variants from an exome sequencing study, it took two and half minutes by ANNOVAR (see below) in my computer. But it would have taken over 20 minutes for tabix in the same computer in my own benchmarking study (not to mention that you would still need to process tabix results file to find the exact matches of nucleotide since tabix operates on region and does not check nucleotide identity).

[kaiwang@biocluster ~/]$ time annotate_variation.pl 84060.avinput humandb/ -dbtype gerp++gt2 -build hg19 -filter
NOTICE: the --dbtype gerp++gt2 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to 84060.avinput.hg19_gerp++gt2_dropped, other variants are written to 84060.avinput.hg19_gerp++gt2_filtered
NOTICE: Processing next batch with 30726 unique variants in 30726 input lines
NOTICE: Database index loaded. Total number of bins is 1978403 and the number of bins to be scanned is 24095
NOTICE: Scanning filter database humandb/hg19_gerp++gt2.txt...Done

real 3m4.598s
user 2m33.670s
sys 0m2.736s

CG (complete genomics) frequency annotations

Each technical platform, such as Complete Genomics and Illumina HiSeq, may generate some platform specific sequencing artifacts. Complete genomics provides whole-genome data for a relatively small group of healthy subjects, but this data set can be quite useful to filter out technical artifacts for CG users. Currently, cg46 and cg69 are provided in ANNOVAR, representing allele frequency data from 46 unrelated subjects and 69 related subjects (including the 46 unrelated subjects), respectivly. Remember that 46 subjects have only 92 autosomes, so even if a mutation is observed once, it will still have MAF>1% in this data set!

[kaiwang@biocluster ~/]$ annotate_variation.pl -downdb -webfrom annovar -build hg19 cg46 humandb/

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -out ex1 -dbtype cg46 -build hg19 example/ex1.avinput humandb/
NOTICE: the --dbtype cg46 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_cg46_dropped, other variants are written to ex1.hg19_cg46_filtered
NOTICE: Processing next batch with 12 unique variants in 12 input lines
NOTICE: Database index loaded. Total number of bins is 2783303 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_cg46.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_cg46_dropped
cg46 0.011 1 13211293 13211294 TC - comments: rs59770105, a 2-bp deletion
cg46 0.620 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
cg46 0.620 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
cg46 0.043 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
cg46 0.511 1 84875173 84875173 C T comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
cg46 0.793 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
cg46 0.022 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
cg46 0.304 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

The commands above first download the "cg46" database to the humandb/ directory, then scan the ex1.avinput file against the database. Variants that are found in cg46, as well as their allele frequencies, are written to the *dropped file:

As you can see, the rs59770105 SNP occured only once in the cg46 database, with allele frequency of 0.011. If you compare the results with the esp6500_ea results, you will see that more SNPs can be identified from cg46 (which is not surprising as cg46 is for whole-genome variants), but the allele frequency measure in esp6500_ea may be more accurate given the larger number of samples from that database.

We can set up a MAF threshold of 0.05, so that only very common SNPs are dropped. We also add -out argument to specify output file name prefix:

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype cg46 -build hg19 example/ex1.avinput humandb/ -score_threshold 0.05 -out ex1
NOTICE: the --dbtype cg46 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_cg46_dropped, other variants are written to ex1.hg19_cg46_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 2783303 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_cg46.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_cg46_dropped
cg46 0.620 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
cg46 0.620 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
cg46 0.511 1 84875173 84875173 C T comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
cg46 0.793 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
cg46 0.304 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

PopFreqMax and PopFreqAll annotations

The popfreq_max database contains the maximum allele frequency from several population frequency databases, including 1000 Genomes Project (ALL+5 ethnicity groups), ESP6500 (ALL+2 ethnicity groups), ExAC (ALL+7 ethnicity groups), CG46. The popfreq_all database contains multiple columns representing all allele frequency measures from these population frequency databases.

The name of the database is suffixed by the date when the database is made. Currently, 20150413 is the suffix.

For example, to check the allele frequency information for variants in the ex1.avinput file:

[kaiwang@biocluster ~/project/annotate_variation]$ annotate_variation.pl -downdb -webfrom annovar -build hg19 popfreq_max_20150413 humandb/

[kaiwang@biocluster ~/project/annotate_variation]$ annotate_variation.pl -filter -out ex1 -dbtype popfreq_max_20150413 -build hg19 example/ex1.avinput humandb/
NOTICE: the --dbtype popfreq_max_20150413 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_popfreq_max_20150413_dropped, other variants are written to ex1.hg19_popfreq_max_20150413_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 26410909 and the number of bins to be scanned is 15
NOTICE: Scanning filter database humandb/hg19_popfreq_max_20150413.txt...Done

[kaiwang@biocluster ~/project/annotate_variation]$ head -n 5 ex1.hg19_popfreq_max_20150413_dropped
popfreq_max_20150413    0.011   1       13211293        13211294        TC      -       comments: rs59770105, a 2-bp deletion
popfreq_max_20150413    0.15    1       1404001 1404001 G       T       comments: rs149123833, a SNP in 3' UTR of ATAD3C
popfreq_max_20150413    0.85    1       162736463       162736463       C       T       comments: rs1000050, a SNP in Illumina SNP arrays
popfreq_max_20150413    0.92    1       5935162 5935162 A       T       comments: rs1287637, a splice site variant in NPHP4
popfreq_max_20150413    0.064   1       67705958        67705958        G       A       comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease

The same procedure can be used for popfreq_all_20150413, though one may want to add -otherinfo argument in the command line so that all frequency information is printed out. This database is more useful when using table_annovar.pl, as the column header will be printed out automatically in the output file.

CLINVAR annotations

The ClinVar database archives and aggregates information about relationships among variation and human health. Accessions, of the format SCV000000000.0, are assigned to each record.

I compiled the database to be used by ANNOVAR. Note that the database name has a date (which is identical to Clinvar release date) as suffix, so make sure to check the download page to use the latest version.

Starting from March 2015, all Clinvar databases in ANNOVAR are processed by the allele-splitting and left-normalization pipeline described here.

An example command is below:

[kaiwang@biocluster ~/]$ annotate_variation.pl example/ex1.avinput humandb/ -filter -dbtype clinvar_20140211 -buildver hg19 -out ex1 
NOTICE: the --dbtype clinvar_20140211 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_clinvar_20140211_dropped, other variants are written to ex1.hg19_clinvar_20140211_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 21199 and the number of bins to be scanned is 6
NOTICE: Scanning filter database humandb/hg19_clinvar_20140211.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19.avinput.hg19_clinvar_20140211_dropped
clinvar_20140211 CLINSIG=other|other;CLNDBN=Inflammatory_bowel_disease_17\x2c_protection_against|Psoriasis\x2c_protection_against;CLNACC=RCV000003254.2|RCV000003255.2 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
clinvar_20140211 CLINSIG=pathogenic|pathogenic|pathogenic|pathogenic;CLNDBN=Deafness\x2c_autosomal_recessive_1A|Deafness\x2c_digenic\x2c_GJB2/GJB6|Hereditary_hearing_loss_and_deafness|not_provided;CLNACC=RCV000018527.27|RCV000018528.27|RCV000037843.1|RCV000080373.1 13 20763686 20763686 comments: rs1801002 (del35G), a frameshift mutation in GJB2, associated with hearing loss
clinvar_20140211 CLINSIG=other;CLNDBN=Inflammatory_bowel_disease_1\x2c_susceptibility_to;CLNACC=RCV000004957.1 16 50745926 50745926 comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
clinvar_20140211 CLINSIG=other;CLNDBN=Inflammatory_bowel_disease_1\x2c_susceptibility_to;CLNACC=RCV000004956.1 16 50756540 50756540 comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
clinvar_20140211 CLINSIG=other;CLNDBN=Inflammatory_bowel_disease_10\x2c_susceptibility_to;CLNACC=RCV000001189.1 2 234183368 234183368 comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's diseasee

SIG refers to Variant Clinical Significance, including unknown, untested, non-pathogenic, probable-non-pathogenic, probable-pathogenic, pathogenic, drug-response,histocompatibility, other. CLINDBN refers to Variant disease name, CLINACC refers to Variant Accession and Versions.

CADD annotations

CADD (Combined Annotation Dependent Depletion) is a score that is based on SVM on multiple other scores. One nice thing about it is that it assigns a score to each possible mutation in the human genome, therefore can evaluate non-coding variants as well as coding ones. The problem with this database is that the size is just too big (350GB in ANNOVAR format). However, several ANNOVAR users requested that I make it available through ANNOVAR, so that they can realistically use CADD in their exome sequencing studies. Therefore, I made the database available in Februray 2014 to ANNOVAR users. Download it by -downdb cadd -buildver hg19, but before you download/use it, make sure that you have the appropriate license to use it (contact the authors themselves to obtain the license). Unfortunately I do not have the resources to make hg18 scores available.

[kaiwang@biocluster ~/]$ annotate_variation.pl example/ex1.avinput humandb/ -filter -dbtype cadd -buildver hg19 -out ex1 
NOTICE: the --dbtype cadd is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_cadd_dropped, other variants are written to ex1.hg19_cadd_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 286120 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_cadd.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_cadd_dropped 
cadd -0.910814 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
cadd 0.515537 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
cadd 0.905064 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
cadd 4.183326 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
cadd 0.008968 1 84875173 84875173 C T comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
cadd 0.175820 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
cadd 3.578490 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
cadd 4.373037 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
cadd 1.504884 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

It takes less than one minute in my computer to obtain CADD scores for these variants, but 99% of the time was spent on loading index. 5GB memory is required to load index into memory. Please note that only SNPs have CADD scores, as indels/CNVs cannot be matched to CADD database.

To know the phred-scaled CADD score, add -otherinfo argument

[kaiwang@biocluster ~/]$ annotate_variation.pl example/ex1.avinput humandb/ -filter -dbtype cadd -buildver hg19 -out ex1 -otherinfo
NOTICE: the --dbtype cadd is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_cadd_dropped, other variants are written to ex1.hg19_cadd_filtered
NOTICE: Processing next batch with 12 unique variants in 12 input lines
NOTICE: Database index loaded. Total number of bins is 286120 and the number of bins to be scanned is 12
NOTICE: Scanning filter database humandb/hg19_cadd.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_cadd_dropped
cadd -0.910814,0.416 1 1404001 1404001 G T comments: rs149123833, a SNP in 3' UTR of ATAD3C
cadd 0.515537,6.794 1 162736463 162736463 C T comments: rs1000050, a SNP in Illumina SNP arrays
cadd 0.905064,8.681 1 5935162 5935162 A T comments: rs1287637, a splice site variant in NPHP4
cadd 4.183326,21.7 1 67705958 67705958 G A comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
cadd 0.008968,4.060 1 84875173 84875173 C T comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
cadd 0.175820,4.946 1 948921 948921 T C comments: rs15842, a SNP in 5' UTR of ISG15
cadd 3.578490,18.23 16 50745926 50745926 C T comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
cadd 4.373037,23.1 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
cadd 1.504884,10.98 2 234183368 234183368 A G comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

The comma-delimited values are raw scores and phred-scaled scores. Basically, for scaled scores, 10 means 10% percentile highest scores, 20 means 1% percentile highest scores, and 30% means 0.1% percentile highest scores. So we do find one <1% percentile variant in the dataset.

I evaluated the performance to annotate ~30,000 variants from an exome sequencing project. It takes 2.5 minutes in my computer.

[kaiwang@biocluster ~/]$ annotate_variation.pl 84060.avinput humandb/ -filter -dbtype cadd -buildver hg19 -out ex1
NOTICE: the --dbtype cadd is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.hg19_cadd_dropped, other variants are written to ex1.hg19_cadd_filtered
NOTICE: Processing next batch with 30726 unique variants in 30726 input lines
NOTICE: Database index loaded. Total number of bins is 14293554 and the number of bins to be scanned is 27608
NOTICE: Scanning filter database tempdb/hg19_cadd.txt...Done

In comparison, it would take 22 minutes in tabix, not to mention the need to post-processing tabix results to filter for nucleotide-level matches. So it seems that ANNOVAR is >8X faster than tabix for this particular data set. Performance-wise, it is practical to use ANNOVAR in at least exome sequencing studies.

DANN annotations

DANN scores whole-genome variants by training a deep neural network (DNN). DNNs can capture non-linear relationships among features and are better suited than SVMs for problems with a large number of samples and features. It seems to perform much better than CADD and FATHMM (http://www.enlis.com/blog/2015/03/17/the-best-variant-prediction-method-that-no-one-is-using/).

Users can use -build hg19 -downdb dann . to download this database with ANNOVAR index files. Note that this file is over 200GB and it may take a while to get it.

Eigen score annotations

Eigen score uses a spectral approach integrating functional genomic annotations for coding and noncoding variants. Unlike other approaches such as CADD/DANN, Eignen did not use labelled training data and should be considered as a unsupervised learning approach. Currently only hg19 is supported

Users can use -build hg19 -downdb eigen . to download this database with ANNOVAR index files. Note that this file is over 200GB and it may take a while to get it. An example is given below:

[kaiwang@dragon ~/]$ annotate_variation.pl -filter ex1.avinput humandb/ -dbtype eigen -build hg19
NOTICE: the --dbtype eigen is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.avinput.hg19_eigen_dropped, other variants are written to ex1.avinput.hg19_eigen_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 26412189 and the number of bins to be scanned is 15
NOTICE: Scanning filter database humandb/hg19_eigen.txt...Done

[kaiwang@dragon ~/]$ cat ex1.avinput.hg19_eigen_dropped
eigen   -1.4191 1       1404001 1404001 G       T       comments: rs149123833, a SNP in 3' UTR of ATAD3C
eigen   0.0309  1       162736463       162736463       C       T       comments: rs1000050, a SNP in Illumina SNP arrays
eigen   0.1277  1       5935162 5935162 A       T       comments: rs1287637, a splice site variant in NPHP4
eigen   0.5929  1       67705958        67705958        G       A       comments: rs11209026 (R381Q), a SNP in IL23R associated with Crohn's disease
eigen   0.0120  1       84875173        84875173        C       T       comments: rs6576700 or SNP_A-1780419, a SNP in Affymetrix SNP arrays
eigen   -0.0445 1       948921  948921  T       C       comments: rs15842, a SNP in 5' UTR of ISG15
eigen   0.1154  16      50745926        50745926        C       T       comments: rs2066844 (R702W), a non-synonymous SNP in NOD2
eigen   0.7733  16      50756540        50756540        G       C       comments: rs2066845 (G908R), a non-synonymous SNP in NOD2
eigen   -1.4621 2       234183368       234183368       A       G       comments: rs2241880 (T300A), a SNP in the ATG16L1 associated with Crohn's disease

COSMIC annotations

COSMIC refers to "Catalogue Of Somatic Mutations In Cancer". It includes somatic mutations reported in literature in various types of cancers. ANNOVAR users can scan a file with somatic muations against the database, to know whether a mutation has previously been reported or observed, and in what types of cancer, and for how many times.

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -build hg19 -out ex4 -dbtype cosmic68 ex4.avinput humandb/
NOTICE: the --dbtype cosmic68 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex4.hg19_cosmic68_dropped, other variants are written to ex4.hg19_cosmic68_filtered
NOTICE: Processing next batch with 16298 unique variants in 16298 input lines
NOTICE: Database index loaded. Total number of bins is 231486 and the number of bins to be scanned is 1738
NOTICE: Scanning filter database humandb/hg19_cosmic68.txt...Done

[kaiwang@biocluster ~/]$ cat ex4.hg19_cosmic68_dropped 
cosmic68 ID=COSM1356066;OCCURENCE=1(large_intestine) 11 6579854 6579854 C T het

The input file contains two variants that were reported in COSMIC database, both reported in cancer involving large inestine only once.

Please note that COSMIC will no longer be updated due to license restrictions. The latest version from ANNOVAR is cosmic70. If you wish to use a more updated version, you will need to acquire COSMIC yourself and use it within ANNOVAR. The general instructions on preparing your own COSMIC data for user in ANNOVAR is given below (using version 76 and GRCh38 coordinate as an example):

  1. Download these VCF files for coding and non-coding mutations: /files/grch38/cosmic/v76/VCF/CosmicCodingMuts.vcf.gz and CosmicNonCodingVariants.vcf.gz from COSMIC's SFTP repository (sftp-cancer.sanger.ac.uk)
  2. Download MutantExport files: /files/grch38/cosmic/v76/CosmicMutantExport.tsv.gz
  3. Use this command prepare_annovar_user.pl -dbtype cosmic CosmicMutantExport.tsv -vcf CosmicCodingMuts.vcf > hg38_cosmic76.txt

Note that the prepare_annovar_user.pl file can be downloaded from here. The final result file should contain coding mutations from COSMIC, as well as the number of occurrences in different tumor types (However, note that these include both targeted screen and genome screen. If you only want genome screen, you should use the CosmicGenomeScreensMutantExport.tsv.gz file instead). COSMIC changed their data formats so non-coding mutations are no longer in the MutantExport file, so we can no longer calculate their occurrences in various tumors. COSMIC now provides a CosmicNCV.tsv file, but it is not really that informative as the cancer tissue information is missing from this file.

ICGC annotations

icgc refers to International Cancer Genome Consortium. The instructions to build the database and the inital database were provided by Ferran Nadeu at Institut d'Investigacions Biomediques August Pi i Sunyer. The annotations include a ICGC_ID column and a ICGC_Occurrence column. The ICGC_Occurrence column includes the project in which the mutation was identified, the number of donors affected, the total number of donors studied in the project and the frequency of the mutation, separated by "|".

[kaiwang@biocluster ~/]$ table_annovar.pl ex1.avinput humandb/ -protocol icgc21 -operation f -build hg19
-----------------------------------------------------------------
NOTICE: Processing operation=f protocol=icgc21
NOTICE: Finished reading 2 column headers for '-dbtype icgc21'

NOTICE: Running system command <annotate_variation.pl -filter -dbtype icgc21 -buildver hg19 -outfile ex1.avinput ex1.avinput humandb/ -otherinfo>
NOTICE: the --dbtype icgc21 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex1.avinput.hg19_icgc21_dropped, other variants are written to ex1.avinput.hg19_icgc21_filtered
NOTICE: Processing next batch with 15 unique variants in 15 input lines
NOTICE: Database index loaded. Total number of bins is 2831634 and the number of bins to be scanned is 15
NOTICE: Scanning filter database humandb/hg19_icgc21.txt...Done
-----------------------------------------------------------------
NOTICE: Multianno output file is written to ex1.avinput.hg19_multianno.txt

Several mutations are present in ICGC and the related information is in the output file.

NCI60 annotations

The NCI60 data set was based on this paper. The NCI-60 cell lines are the most frequently studied human tumor cell lines in cancer research. I compiled allele frequency information from these 60 cell lines based on their exome sequencing data. Example usage is below, using the 'nci60' as the keyword.

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -build hg19 -out ex4 -dbtype nci60 ex4.avinput humandb/
NOTICE: the --dbtype nci60 is assumed to be in generic ANNOVAR database format
NOTICE: Variants matching filtering criteria are written to ex4.hg19_nci60_dropped, other variants are written to ex4.hg19_nci60_filtered
NOTICE: Processing next batch with 16298 unique variants in 16298 input lines
NOTICE: Database index loaded. Total number of bins is 81308 and the number of bins to be scanned is 766
NOTICE: Scanning filter database humandb/hg19_nci60.txt...Done

Generic mutation annotations

Users have the flexibility to supply a custom-made annotation file, and let ANNOVAR perform filter-based annotation on this annotation file. The -dbtype should be specified as "generic". The file format is very simple: first five columns are chr, start, end, reference allele, observed allele, the sixth column (functional score) is optional, other columns are optional as well asn will be ignored. The first ten lines of an example database file is given below:

[kai@biocluster ~/]$ head example/hg19_example_db_generic.txt 
16 50756539 50756539 T A 1
16 50756539 50756539 T C 1
16 50756539 50756539 T G 0.04
16 50756540 50756540 G A 0.04
16 50756540 50756540 G C 0.05
16 50756540 50756540 G T 0.05
16 50756541 50756541 G A 0.05
16 50756541 50756541 G C 0.01
16 50756541 50756541 G T 0.01
16 50756542 50756542 G A 0.71

[kaiwang@biocluster ~/]$ annotate_variation.pl -filter -dbtype generic -genericdbfile hg19_example_db_generic.txt -build hg19 -out ex1 example/ex1.avinput example/
NOTICE: Variants matching filtering criteria are written to ex1.hg19_generic_dropped, other variants are written to ex1.hg19_generic_filtered
NOTICE: Processing next batch with 12 unique variants in 12 input lines
NOTICE: Scanning filter database example/hg19_example_db_generic.txt...Done

[kaiwang@biocluster ~/]$ cat ex1.hg19_generic_dropped
generic 0.05 16 50756540 50756540 G C comments: rs2066845 (G908R), a non-synonymous SNP in NOD2

Running the above command, we can see that one variant is present in the generic database hg19_example_db_generic.txt, and its function score (as 0.05) is printed as the second column in the output file ex1.hg19_generic_dropped.

When the database file contains the functional importance score column (sixth column in the generic format files), users can use --score_threshold argument to control output: only database records with a higher functional score will be used in filtering query variants. For example, when adding -score_threshold 0.5 argument to the above command, the output file will be empty.

It is also possible to treat ljb23_pp2, ljb23_lrt, etc as if they are generic database file, and use the -dbtype generic -genericdbfile hg19_ljb23_pp2hvar.txt argument to perform PolyPhen annotation of variants.

It is also possible to make two files that each contains a list of variants (five columns in each file), then scan one file against the other, to find the shared variants. This is useful, for example, when comparing two genotype-calling algorithms on the same set of alignment data.