clintval / vartovcf / 37253647582

98%

Build Type

push

github

Committed by web-flow

Commit Message

feat!: support VarDictJava tumor-normal output (#133)

* feat!: detect the input layout and sample from the first row and drop --mode

vartovcf now peeks at the first row to tell VarDictJava's tumor-only rows (segment in column 35) from its tumor-normal rows (segment in column 53), and refuses either layout without the --fisher columns. --mode is removed.

--sample is now optional: the tumor sample's name is read from the first row, and when --sample is given it must match. VarDictJava never writes the normal sample's name, so the new --normal-sample is required for tumor-normal input and refused for tumor-only input. With empty input there is nothing to peek at, so --sample (and --normal-sample, for a two-sample header) name the sample columns, and neither given is an error. Tumor-normal rows are refused for now, until the tumor-normal writer lands.

* feat!: give a call with no ALT reads no ALT-read statistics

Means over the ALT reads (QMEAN, MEAN_MAPQ, MEAN_MISMATCHES, MEAN_DIST_TO_READ_END, ALT_READ_POS_VARIES, STRAND_BIAS_FISHER_P, REALIGNED_FRAC_OF_DP) are now missing when no read carries the ALT allele, rather than VarDict's zeros or its REF-read values. GT is 0/0 without ALT reads and ./. when DP is 0 too. FILTER labels test the FORMAT values, so a missing value never applies a label. A tumor-normal row often carries a sample with no reads of the allele.

* feat: add a one-sided Fisher exact test

* feat!: write tumor-normal VCFs

Tumor-normal rows now convert to records with a tumor and a normal sample column, carrying the tumor-only FORMAT fields except HICNT, which VarDict does not write per sample. INFO VARDICT_STATUS passes VarDict's tumor-normal label through with a description of what each label really tests, and INFO TUMOR_NORMAL_FISHER_P replaces VarDict's SSF and SOR with a one-sided Fisher exact p-value that the allele makes up more of the tumor's reads than of the normal's. FILTER labels test the tumor sample, with no rescue of ca... (continued)

Coverage Stats

585 of 593 new or added lines in 5 files covered. (98.65%)

2 existing lines in 1 file now uncovered.

1510 of 1528 relevant lines covered (98.82%)

86.65 hits per line