This document provides detailed guidelines for using SNPEff variant annotation tool. --- ## Installation ### Conda Installation (Recommended) ``` # Install SNPEff conda install -c bioconda snpeff # Verify installation snpEff -version ``` ### Manual Installation ``` # Download latest version wget https://snpeff.blob.core.windows.net/versions/snpEff_latest_core.zip # Extract unzip snpEff_latest_core.zip # Test java -jar snpEff/snpEff.jar -version ``` --- ## Database Setup SNPEff requires genome databases for annotation. ### List Available Databases ``` snpEff databases | grep -i human snpEff databases | grep -i GRCh38 ``` Common human databases: - `GRCh38.105`: Ensembl release 105, GRCh38 - `GRCh37.75`: Ensembl release 75, GRCh37 - `hg38`: UCSC hg38 - `hg19`: UCSC hg19 ### Download Database ``` # Download GRCh38 snpEff download GRCh38.105 # Download GRCh37 snpEff download GRCh37.75 # Specify custom data directory snpEff download -dataDir /path/to/data GRCh38.105 ``` ### Database Disk Space - Human genome: ~2-3 GB - Mouse genome: ~1-2 GB --- ## Common Usage Patterns ### Basic Annotation ``` snpEff ann GRCh38.105 input.vcf > output.vcf ``` ### With Statistics ``` snpEff ann \ -stats output_stats.html \ GRCh38.105 \ input.vcf > output.vcf ``` ### Recommended Clinical Configuration ``` snpEff ann \ -dataDir ~/.snpeff \ -stats snpeff_stats.html \ -csvStats snpeff_stats.csv \ -canon \ -hgvs \ -lof \ -no-downstream \ -no-upstream \ -no-intergenic \ GRCh38.105 \ input.vcf > output.vcf ``` --- ## Key Parameters ### Output and Statistics - `-stats file.html`: Generate comprehensive HTML statistics report with plots showing variant distribution, effects by impact, functional class, and genomic region. **Recommended for QC.** - `-csvStats file.csv`: Generate CSV format statistics for programmatic analysis - `-v`: Verbose mode - prints detailed progress and warnings - `-q`: Quiet mode - suppress all output except errors ### Annotation Options - `-canon`: Use only canonical transcripts (one representative transcript per gene). **Recommended for clinical reporting** to reduce redundancy and focus on primary isoforms. - `-hgvs`: Add HGVS nomenclature (e.g., "c.123A>G", "p.Lys41Arg"). **Essential for clinical reporting** as it provides standardized variant descriptions. - `-lof`: Predict high-confidence loss-of-function variants using stringent criteria (stop-gained, frameshift affecting >50% of protein, splice-disrupting, initiator codon variants). Adds LOF and NMD (nonsense-mediated decay) tags. - `-oicr`: Use OICR tag format instead of default ANN format - `-sequenceOntology`: Use Sequence Ontology terms for consequence types (default, recommended) - `-formatEff`: Use old EFF format (deprecated, avoid unless required for legacy tools) ### Filtering Options These parameters reduce output to relevant variants by excluding low-information regions: - `-no-downstream`: Don't annotate downstream gene variants (typically 5kb after gene end). Reduces noise for coding-focused analysis. - `-no-upstream`: Don't annotate upstream gene variants (typically 5kb before gene start). Useful for exome/targeted sequencing. - `-no-intergenic`: Don't annotate intergenic variants (variants between genes). **Recommended for exome analysis** where intergenic variants are not of interest. - `-no-intron`: Don't annotate intronic variants. Use cautiously - some intronic variants affect splicing. - `-no-utr`: Don't annotate UTR (untranslated region) variants. May miss regulatory variants. **Typical exome configuration:** `-no-downstream -no-upstream -no-intergenic` (keeps coding and regulatory variants) ### Performance - `-t N`: Use N threads for parallel processing. Note: Multi-threading support varies by operation; most benefit is during database loading. ### Data Directory - `-dataDir /path/to/data`: Specify custom location for SNPEff databases (default: `~/.snpeff/data/`) ### Input/Output Formats - **Input formats:** VCF (default), BED - **Output formats:** VCF (default), GATK, BED, TXT ### Cancer-Specific - `-cancer`: Enable cancer-specific annotations - `-cancerSamples file.txt`: Specify cancer sample mapping file --- ## ANN Field Format SNPEff adds annotations to the INFO field with the ANN key: ``` ##INFO= ``` ### Example ANN Entry ``` T|missense_variant|MODERATE|BRCA1|ENSG00000012048|transcript|ENST00000357654|protein_coding|10/23|c.1234A>T|p.Lys412Tyr|1234/5592|1234/5643|412/1863|| ``` ### Field Positions 1. Allele 2. Annotation (consequence) 3. Annotation\_Impact (HIGH, MODERATE, LOW, MODIFIER) 4. Gene\_Name 5. Gene\_ID 6. Feature\_Type 7. Feature\_ID (transcript ID) 8. Transcript\_BioType 9. Rank (exon/intron rank) 10. HGVS.c (coding sequence) 11. HGVS.p (protein sequence) 12. cDNA\_position/length 13. CDS\_position/length 14. Protein\_position/length 15. Distance to feature 16. Errors/Warnings/Info --- ## SnpSift for Additional Annotations SnpSift is a companion tool for filtering and annotating VCF files. ### Add dbNSFP Annotations ``` SnpSift dbnsfp \ -db /path/to/dbNSFP4.4a.txt.gz \ -f SIFT_pred,Polyphen2_HVAR_pred,CADD_phred,REVEL_score \ input.vcf > annotated.vcf ``` ### Add ClinVar Annotations ``` SnpSift annotate \ ClinVar.vcf.gz \ input.vcf > annotated.vcf ``` ### Filter Variants ``` # High impact variants only SnpSift filter "(ANN[*].IMPACT = 'HIGH')" input.vcf > high_impact.vcf # Rare variants (gnomAD AF < 0.01) SnpSift filter "(dbNSFP_gnomAD_genomes_AF < 0.01)" input.vcf > rare.vcf ``` --- ## Best Practices ### 1. Use Canonical Transcripts For clinical reporting: ``` snpEff ann -canon GRCh38.105 input.vcf > output.vcf ``` ### 2. Add HGVS Nomenclature Essential for clinical reporting: ``` snpEff ann -hgvs GRCh38.105 input.vcf > output.vcf ``` ### 3. Enable LOF Prediction Identifies high-confidence loss-of-function variants: ``` snpEff ann -lof GRCh38.105 input.vcf > output.vcf ``` LOF criteria: - Stop-gained variants - Frameshift variants affecting >50% of protein - Splice-site disrupting variants - Initiator codon variants ### 4. Filter Non-Coding Variants for Exome ``` snpEff ann \ -no-downstream \ -no-upstream \ -no-intergenic \ GRCh38.105 \ input.vcf > output.vcf ``` ### 5. Generate Statistics Report Always generate HTML report for QC: ``` snpEff ann -stats report.html GRCh38.105 input.vcf > output.vcf ``` Report includes: - Variant type distribution - Effects by impact - Effects by functional class - Effects by genomic region - Base changes - Quality metrics ### 6. Combine with SnpSift ``` # Annotate with SNPEff snpEff ann -hgvs -lof GRCh38.105 input.vcf | \ # Add dbNSFP scores SnpSift dbnsfp -db dbNSFP.txt.gz \ -f SIFT_pred,Polyphen2_HVAR_pred,CADD_phred | \ # Filter high impact SnpSift filter "(ANN[*].IMPACT = 'HIGH') | (ANN[*].IMPACT = 'MODERATE')" \ > filtered.vcf ``` ### 7. Match Database Version to VCF Ensure genome build compatibility: - VCF with GRCh38 → Use GRCh38. *database - VCF with hg19/GRCh37 → Use GRCh37.* database --- ## Cancer Analysis with SNPEff ### Annotate Somatic Variants ``` snpEff ann \ -cancer \ -cancerSamples cancer_samples.txt \ GRCh38.105 \ tumor.vcf > tumor.annotated.vcf ``` ### Add COSMIC Annotations ``` SnpSift annotate \ CosmicCodingMuts.vcf.gz \ input.vcf > cosmic_annotated.vcf ``` --- ## Troubleshooting ### "Cannot find database" **Cause:** Database not downloaded **Solution:** ``` # List available databases snpEff databases | grep GRCh38 # Download database snpEff download GRCh38.105 ``` ### "Chromosome not found" **Cause:** Chromosome naming mismatch (chr1 vs 1) **Solution:** - Check VCF chromosome names: `grep -v "^#" input.vcf | cut -f1 | sort -u` - SNPEff handles both "chr1" and "1" formats automatically - If issues persist, use alternative database (e.g., hg38 instead of GRCh38) ### Empty ANN Field **Cause:** No overlapping annotations or genome build mismatch **Solution:** - Verify genome build matches VCF - Check that variants are in coding regions - Use `-v` for verbose output to debug ### Very Large Output Files **Cause:** Annotating all transcripts and intergenic regions **Solution:** ``` # Filter to coding variants only snpEff ann \ -no-intergenic \ -no-downstream \ -no-upstream \ GRCh38.105 input.vcf > output.vcf ``` ### Statistics Show Low Annotation Rate **Causes:** - Genome build mismatch - Poor quality variants - Non-coding variants (expected for WGS) **Investigation:** ``` # Check variant distribution in stats HTML firefox snpeff_stats.html # Check example variants head -n 1000 output.vcf | grep "ANN=" ``` --- ## Performance Optimization ### For Large VCFs ``` # Process by chromosome for chr in {1..22} X Y; do grep "^${chr}\t" input.vcf | snpEff ann GRCh38.105 /dev/stdin > chr${chr}.vcf done # Merge results bcftools concat chr*.vcf -o merged.vcf ``` ### Memory Settings ``` # Increase Java heap size (default: 1GB) java -Xmx8g -jar snpEff.jar ann GRCh38.105 input.vcf > output.vcf ``` --- ## Comparison with VEP | Feature | SNPEff | VEP | | --- | --- | --- | | Speed | Faster | Slower (Perl) | | Organisms | 38,000+ | ~100 well-annotated | | Installation | Simpler | More complex | | Output Format | ANN field (standard) | CSQ field | | Database Size | Smaller (2-3 GB) | Larger (15-20 GB) | | Clinical Databases | Via SnpSift | Native integration | | Plugins | Limited | Extensive | | Updates | Less frequent | Quarterly | | GATK Integration | Excellent | Good | **Use SNPEff when:** - Working with non-model organisms - Speed is priority - Simple setup needed - GATK pipeline integration **Use VEP when:** - Comprehensive human clinical annotation - Need latest databases - Extensive plugin ecosystem required --- ## References - SNPEff Documentation: https://pcingola.github.io/SnpEff/ - SNPEff Paper: Cingolani P, et al. (2012) Fly 6(2):80-92 - SnpSift Documentation: https://pcingola.github.io/SnpEff/ss\_introduction/