broadinstitute / picard

A set of command line tools (in Java) for manipulating high-throughput sequencing (HTS) data and formats such as SAM/BAM/CRAM and VCF.
https://broadinstitute.github.io/picard/
MIT License
959 stars 366 forks source link

ValidateSamFile wrong NM tag computation #1963

Open albertocasagrande opened 1 month ago

albertocasagrande commented 1 month ago

Bug Report

Affected tool(s)

picard ValidateSameFile

Affected version(s)

Description

Calling picard ValidateSamFile with a reference sometimes produces ERROR:INVALID_TAG_NM even though the read alignment on the reference and CIGAR code match.

Steps to reproduce

  1. Download and decompress the reference sequence
    curl https://ftp.ensembl.org/pub/grch37/release-111/fasta/homo_sapiens/dna/Homo_sapiens.GRCh37.dna.chromosome.22.fa.gz -o chr22.fasta.gz
    gunzip chr22.fasta.gz
  2. Create the file test.sam containing the following lines
    @HD VN:1.6  SO:unknown
    @SQ SN:22   LN:51304566 AN:chromosome22,chr22,chromosome_22,chr_22
    @RG ID:S_1_1    SM:S_1_1    PL:ILLUMINA
    r00000000001    16  22  42063771    33  55M2D95M    *   0   0   CTTCTAGTGTTCCTCGCTACCCTGCAATTTTAGCATGACCATTTATTTATTTATGTTTGTTTGTTTATTTATTTATTTATGACTACAAAGATCCAGAAGACAAACATGACCATTTCTTTCTTTTTTTTTTTTTTTCTGAGATGGAGTCTT  !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!  RG:Z:S_1_1  NM:i:2
    r00000000002    16  22  50364657    33  121M29X *   0   0   GTGACAGGGGAGGAGTCTGGAGCTGAGAGGCGAACGGAGAGCACAGTGGAGCACACGGGCCCTGCCCACCCGCCTGTCCTGTCCAAGGATGCTGGGGCCCCGACCAGCCGGTCACAGGCGCNNNNNNNNNNNNNNNNNNNNNNNNNNNNN  !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!  RG:Z:S_1_1  NM:i:29
  3. Validate the sam file with Picard
    java -jar picard.jar ValidateSamFile -I test.sam -R chr22.fasta

Expected behavior

The validation should succeed. The first read alignment and its CIGAR correspond. The second read perfectly matches the reference, but the final 29 'N's should be considered mismatches according to the NM specification given in "Sequence Alignment/Map Optional Fields Specification" (see at the bottom of p.3).

Actual behavior

Picard produces the following INVALID_TAG_NM errors:

ERROR::INVALID_TAG_NM:Record 1, Read name r00000000001, NM tag (nucleotide differences) in file [2] does not match reality [3]
ERROR::INVALID_TAG_NM:Record 2, Read name r00000000002, NM tag (nucleotide differences) in file [29] does not match reality [0]
droazen commented 1 month ago

The tool calls SequenceUtil.calculateSamNmTag() in HTSJDK to calculate the "expected" NM tag. That method internally calls a helper method SequenceUtil.countMismatches() with the matchAmbiguousRef argument set false. The matchAmbiguousRef toggle controls whether to do strict base matching, or match subsets of possible bases in the case of ambiguity codes, with N defined as an ambiguity code that matches any base.

It seems to me that this matchAmbiguousRef toggle in SequenceUtil.countMismatches() might be what we need here. We could expose it as an additional argument in a new overload of SequenceUtil.calculateSamNmTag() (and possibly as an argument to the ValidateSamFile tool itself).

Note that although the SAM spec does state that "ambiguity codes in both sequence and reference that match each other, such as ‘N’ in both, or compatible codes such as ‘A’ and ‘R’, are still counted as mismatches", the spec also acknowledges that "historically this has been ill-defined and both data and tools exist that disagree with this definition."

kockan commented 1 month ago

The problem seems to be that if both the reference and read base is N, no matter what matchAmbiguousRef is set to, htsjdk will accept this as a match. It obviously should if matchAmbiguousRef is set to true, but the default is false and the code jumps here: https://github.com/samtools/htsjdk/blob/127f3de750acb6a59fe1ceb5865ebbaee938847f/src/main/java/htsjdk/samtools/util/SequenceUtil.java#L132 where a simple equality check is performed, resulting in a match.

One potential idea (in case this is really the issue) might be to add something like https://github.com/samtools/htsjdk/blob/127f3de750acb6a59fe1ceb5865ebbaee938847f/src/main/java/htsjdk/samtools/util/SequenceUtil.java#L184 (but without the N) to only allow certain bases with matchAmbiguousRef set to false.

kockan commented 1 month ago

This was exactly the issue and I thought I had a fix for it, only to discover @michaelgatzen already has something in place from a while back: https://github.com/samtools/htsjdk/pull/1536 Opinions @droazen @lbergelson ?