This is an archive page for SAGE v1.1 which is no longer supported.
SAGE is a somatic SNV, MNV and small INDEL caller optimised to call narrow regions of the genome with high prior chance of a variant with very high sensitivity. SAGE looks directly at the provided reference and tumor bams looking for evidence of:
- Inframe indels within the specified coding regions; and
- Specific hotspots at known locations.
All sites examined are written to VCF with some soft filtering applied to the results.
The supplied set of known hotspots is combined with any inframe indels found within specified regions of the tumor to form the complete set of sites for which to collect evidence.
To find inframe indel sites, SAGE examines the CIGAR field of all tumor alignments within the supplied coding regions looking for inserts or deletes that are divisible by 3, e.g., 101M3D50M and 60M9I82M.
In an alignment, evidence of the alt is accumulated only if:
- The unmapped, duplicated, secondary and supplementary flags are not set.
- The quality of the mapping is sufficient [1];
- Quality of the base is sufficient [13]; and
- The alt matches exactly.
The base quality of an insert, SNV or MNV in a single alignment is calculated as the average base quality of the alt bases. The base quality of a delete is taken from the base after the delete if available, otherwise from the base prior. The minimum acceptable quality is set with min_base_quality [13].
SNVs and MNVs cannot be part of a larger MNV, eg, CA > TT does not match CAT > TTG. There must be at least one base either side of the variant that match the reference exactly.
Only simple indels are considered (eg A > ACT). Evidence of complex indels (eg, A > CCT) will not accumulate.
When writing to a file, a number of soft filters are applied. The quality filters are applied to the sum of the variant base quality scores.
Any variant will be filtered as LOW_CONFIDENCE under the following conditions:
- Insufficient tumor allelic depth, i.e., tumor allelic depth < min_tumor_reads [2]; or
- Evidence of support for the ALT in the germline sample, ie. filter if germline ALT read count >1 OR if germline ALT read count = 1 AND germline heterozygous binomial likelihood > max_het_binomial_likelihood [0.01]
Any snv or mnv will be filtered as LOW_CONFIDENCE if:
- Insufficient quality, i.e., quality < min_snv_quality [100]; or
- Insufficient VAF, i.e., vaf < min_snv_vaf [0.005]
Any indel will be filtered as LOW_CONFIDENCE if:
- Insufficient quality, i.e., quality < min_indel_quality [150]; or
- Insufficient VAF, i.e., vaf < min_indel_vaf [0.02]
Any unfiltered indel will be filtered as GERMLINE_INDEL if there exists an indel at that site in the germline.
Only one variant per site is written to file as ranked by filter then quality.
A snippet of output follows:
##fileformat=VCFv4.2
##FILTER=<ID=GERMLINE_INDEL,Description="Set if indel has any germline indels at that site.">
##FILTER=<ID=LOW_CONFIDENCE,Description="Set if excessive germline reads or insufficient quality or tumor reads">
##FILTER=<ID=PASS,Description="All filters passed">
##FORMAT=<ID=AD,Number=R,Type=Integer,Description="Allelic Depth">
##FORMAT=<ID=DP,Number=1,Type=Integer,Description="Read Depth">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##INFO=<ID=AF,Number=A,Type=Float,Description="Allelic Frequency">
##INFO=<ID=GHBL,Number=1,Type=Float,Description="Germline Heterozygous Binomial Likelihood. Only applies when germline alt support = 1">
##INFO=<ID=HOTSPOT,Number=1,Type=String,Description="Hotspot Type: known, inframe">
#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT COLO829R COLO829T
1 9780851 . GAG AAA 0 LOW_CONFIDENCE AF=0.00;HOTSPOT=known GT:AD:DP 0/1:31,0:32 0/1:56,0:58
5 1295228 . GG AA 1113 PASS AF=0.917;HOTSPOT=known GT:AD:DP 0/1:17,0:20 0/1:0,33:36
7 128829039 . GGCT G 203 PASS AF=0.055;HOTSPOT=inframe GT:AD:DP 0/1:36,0:36 0/1:103,6:109
7 140453136 . A T 4047 PASS AF=0.642;HOTSPOT=known GT:AD:DP 0/1:35,0:35 0/1:56,104:162
11 1018167 . ATCG A 253 GERMLINE_INDEL AF=0.051;GHBL=0.00;HOTSPOT=inframe GT:AD:DP 0/1:71,1:72 0/1:148,8:156
| Argument | Description |
|---|---|
| coding_regions | Coding regions bed file to search for inframe indels |
| known_hotspots | Tab separated file of known hotspot locations |
| max_het_binomial_likelihood | Up to 1 read of support for the ALT in the reference is allowed if the binomial likelihood of observing the read (assuming a 50% VAF) is less than this parameter [0.01] |
| min_base_quality | Minimum quality for a base to be considered [13] |
| min_indel_quality | Low confidence filtering minimum indel quality [150] |
| min_indel_vaf | Low confidence filtering minimum indel VAF [0.02] |
| min_mapping_quality | Minimum mapping quality for an alignment to be used [1] |
| min_snv_quality | Low confidence filtering minimum SNV/MNV quality [100] |
| min_snv_vaf | Low confidence filtering minimum SNV/MNV VAF [0.005] |
| min_tumor_reads | Low confidence filtering minimum tumor reads [2] |
| out | Output VCF file to write. Gz supported |
| ref_genome | Path to the ref genome fasta file |
| reference | Name of reference sample |
| reference_bam | Path to reference bam file |
| tumor | Name of tumor sample |
| tumor_bam | Path to tumor bam file |
java -Xmx8G -Xms4G \
-cp sage.jar com.hartwig.hmftools.sage.SageHotspotApplication \
-tumor COLO829T -tumor_bam /path/to/COLO829T.bam \
-reference COLO829R -reference_bam /path/to/COLO829R.bam \
-known_hotspots /path/to/KnownHotspots.tsv \
-coding_regions /path/to/CodingRegions.bed \
-ref_genome /path/to/refGenome.fasta \
-out COLO829T.vcf.gz
Hotspots file should be header-less and tab separated with columns: Chromosome, Position, Ref, Alt, ie:
1 8073509 C A
1 9777666 C A
1 9777666 C G
1 9780851 G A
The HMF pipeline employs some post processing steps to further filter the output. We have generated a PON file from 1077 sage results that captures any variants that have more than one read in the germline of at least 2 samples. These variants are filtered from our output.
- 1.1
- SageHotspotAnnotation - Merge all info header fields from hotspots vcf with source vcf