NanoSV with truvari

[Tintest]

Hello, I'm trying to benchmark NIST002 vcf files produced by NanoSV against the GIAB NIST002 SV truth set, but not any variant is classfied as true positive.

Here is the logs :

2019-11-20 12:09:19,348 [INFO] Running /home/tintest/miniconda3/bin/truvari -b ../HG002_SVs_Tier1_v0.6.vcf.gz -c /home/tintest/bettik/SV/nanopore/Biomnis/vcf/nanosv/NIST-002_merged_ngmlr_sorted.vcf.gz -o NIST-002_merged_ngmlr_sorted -r 2000 --pctsim 0 --passonly --includebed ../HG002_SVs_Tier1_v0.6.bed --giabreport
2019-11-20 12:09:19,348 [INFO] Params:
{
    "sizemax": 50000,
    "reference": null,
    "noprog": false,
    "multimatch": false,
    "pctsize": 0.7,
    "cSample": null,
    "includebed": "../HG002_SVs_Tier1_v0.6.bed",
    "no_ref": false,
    "passonly": true,
    "pctsim": 0.0,
    "pctovl": 0.0,
    "comp": "/home/tintest/bettik/SV/nanopore/Biomnis/vcf/nanosv/NIST-002_merged_ngmlr_sorted.vcf.gz",
    "refdist": 2000,
    "base": "../HG002_SVs_Tier1_v0.6.vcf.gz",
    "giabreport": true,
    "sizefilt": 30,
    "typeignore": false,
    "gtcomp": false,
    "debug": false,
    "output": "NIST-002_merged_ngmlr_sorted",
    "bSample": null,
    "sizemin": 50
}
2019-11-20 12:09:20,646 [INFO] Including 34830 bed regions
2019-11-20 12:09:20,646 [INFO] Creating call interval tree for overlap search
2019-11-20 12:09:30,717 [INFO] 35145 call variants in total
2019-11-20 12:09:30,717 [INFO] 0 call variants within size range (30, 50000)
2019-11-20 12:09:49,403 [INFO] 20041 base variants
2019-11-20 12:09:49,423 [INFO] Matching base to calls
2019-11-20 12:10:09,183 [WARNING] No TP or FP calls in base!
2019-11-20 12:10:09,183 [INFO] Parsing FPs from calls
2019-11-20 12:10:18,714 [INFO] Stats: {
    "TP-base": 0,
    "TP-call": 0,
    "FP": 0,
    "FN": 9641,
    "precision": 0,
    "recall": 0,
    "f1": "NaN",
    "base cnt": 9641,
    "call cnt": 0,
    "base size filtered": 6309,
    "call size filtered": 0,
    "base gt filtered": 0,
    "call gt filtered": 0,
    "TP-call_TP-gt": 0,
    "TP-call_FP-gt": 0,
    "TP-base_TP-gt": 0,
    "TP-base_FP-gt": 0,
    "gt_precision": 0,
    "gt_recall": 0,
    "gt_f1": "NaN"
}
2019-11-20 12:10:18,715 [INFO] Creating GIAB report
2019-11-20 12:10:20,919 [INFO] Finished

Did you ever tried truvari with a vcf from NanoSV ?

It works flawlessly with SV callers such as Svim, Pbsv or Sniffles.

Here is the link to the vcf : https://filesender.renater.fr/?s=download&token=8905688b-e98a-859c-c841-ad7e9088a2c6

Here is the NanoSV command to produce the vcf file from a bam produced by ngmlr :

singularity exec -B /bettik/tintest/:/mnt /home/tintest/bettik/SV/nanopore/Chaissonetal2019/nanosv.simg NanoSV --bed /mnt/SV/nanopore/human_hg19.bed -t 4 -s samtools /mnt/SV/nanopore/Biomnis/bam/ngmlr/NIST-002_merged_ngmlr.bam -o /mnt/SV/nanopore/Biomnis/vcf/nanosv/NIST-002_merged_ngmlr.vcf

Same problem with NanoSV vcf from bam files produced by minimap2.

I guess the problem must come from the NanoSV vcf format.

Regards.

[yekaizhou]

I came with the same issue, that Truvari detected no SV within size range

[LYC-vio]

you may need to remove the --passonly flag (since NanoSV does not use PASS in vcf's FILTER field), also, you need to change RT=3 to RT=. (or RT=1) in the vcf header

[LYC-vio]

And change ##FILTER=<ID=Gap to ##FILTER=<ID=GAP