midsv

midsv is a Python module that converts SAM files to MIDSV format.

MIDSV (Match, Insertion, Deletion, Substitution, and inVersion) is a comma-separated format that represents differences between a reference and a query, with the same length as the reference.

[!CAUTION] MIDSV is intended for targeted amplicon sequences (10-100 kbp).
Using whole chromosomes as references may exhaust memory and crash.

[!IMPORTANT] MIDSV requires long-format cstag tags in the SAM file.
Please use minimap2 with --cs=long option. or use cstag tool to append long-format cstag.

The output includes MIDSV and, optionally, QSCORE.

• MIDSV preserves original nucleotides while annotating mutations.
• QSCORE provides Phred quality scores for each nucleotide.

Details of MIDSV (formerly MIDS) are described in our paper.

🛠️Installation

From Bioconda (recommended):

conda install -c bioconda midsv

From PyPI:

pip install midsv

📜Specifications

MIDSV

Op	Regex	Description
=	[ACGTN]	Identical sequence
+	[ACGTN]	Insertion to the reference
-	[ACGTN]	Deletion from the reference
*	[ACGTN][ACGTN]	Substitution
	[acgtn]	Inversion
|		Separator for insertion sites

MIDSV uses | to separate nucleotides in insertion sites so +A|+C|+G|+T|=A can be easily split into [+A, +C, +G, +T, =A] by "+A|+C|+G|+T|=A".split("|").

QSCORE

Op	Description
-1	Unknown
|	Separator for insertion sites

QSCORE uses -1 for deletions or unknown nucleotides.

As with MIDSV, QSCORE uses | to separate quality scores in insertion sites.

📘Usage

midsv.transform(
    path_sam: str | Path,
    qscore: bool = False,
    keep: str | list[str] = None
) -> list[dict[str, str | int]]

• path_sam: Path to a SAM file on disk.

• qscore (bool, optional): Output QSCORE. Defaults to False.

• keep: Subset of {‘FLAG’, ‘POS’, ‘SEQ’, ‘QUAL’, ‘CIGAR’, ‘CSTAG’} to include from the SAM file. Defaults to None.

• midsv.transform() returns a list of dictionaries containing QNAME, RNAME, MIDSV, and optionally QSCORE, plus any fields specified by keep.

• MIDSV and QSCORE are comma-separated strings and have the same reference sequence length.

🖍️Examples

Perfect match

import midsv
from midsv.io import read_sam

# Perfect match

path_sam = "examples/example_match.sam"
print(list(read_sam(path_sam)))
# sam = [
#     ['@SQ', 'SN:example', 'LN:10'],
#     ['match', '0', 'example', '1', '60', '10M', '*', '0', '0', 'ACGTACGTAC', '0123456789', 'cs:Z:=ACGTACGTAC']
# ]

print(midsv.transform(path_sam, qscore=True))
# [{
#   'QNAME': 'control',
#   'RNAME': 'example',
#   'MIDSV': '=A,=C,=G,=T,=A,=C,=G,=T,=A,=C',
#   'QSCORE': '15,16,17,18,19,20,21,22,23,24'
# }]

Insertion, deletion, and substitution

import midsv
from midsv.io import read_sam

path_sam = "examples/example_indels.sam"
print(list(read_sam(path_sam)))
# [
#     ['@SQ', 'SN:example', 'LN:10'],
#     ['indel_sub', '0', 'example', '1', '60', '5M3I1M2D2M', '*', '0', '0', 'ACGTGTTTCGT', '01234!!!56789', 'cs:Z:=ACGT*ag+ttt=C-aa=GT']
# ]

print(midsv.transform(path_sam, qscore=True))
# [{
#   'QNAME': 'indel_sub',
#   'RNAME': 'example',
#   'MIDSV': '=A,=C,=G,=T,*AG,+T|+T|+T|=C,-A,-A,=G,=T',
#   'QSCORE': '15,16,17,18,19,0|0|0|20,-1,-1,21,22'
# }]

Large deletion

import midsv
from midsv.io import read_sam

path_sam = "examples/example_large_deletion.sam"
print(list(read_sam(path_sam)))
# [
#     ['@SQ', 'SN:example', 'LN:10'],
#     ['large-deletion', '0', 'example', '1', '60', '2M', '*', '0', '0', 'AC', '01', 'cs:Z:=AC'],
#     ['large-deletion', '0', 'example', '9', '60', '2M', '*', '0', '0', 'AC', '89', 'cs:Z:=AC']
# ]

print(midsv.transform(path_sam, qscore=True))
# [
#   {'QNAME': 'large-deletion',
#   'RNAME': 'example',
#   'MIDSV': '=A,=C,=N,=N,=N,=N,=N,=N,=A,=C',
#   'QSCORE': '15,16,-1,-1,-1,-1,-1,-1,23,24'}
# ]

Inversion

import midsv
from midsv.io import read_sam

path_sam = "examples/example_inversion.sam"
print(list(read_sam(path_sam)))
# [
#     ['@SQ', 'SN:example', 'LN:10'],
#     ['inversion', '0', 'example', '1', '60', '5M', '*', '0', '0', 'ACGTA', '01234', 'cs:Z:=ACGTA'],
#     ['inversion', '16', 'example', '6', '60', '3M', '*', '0', '0', 'CGT', '567', 'cs:Z:=CGT'],
#     ['inversion', '2048', 'example', '9', '60', '2M', '*', '0', '0', 'AC', '89', 'cs:Z:=AC']
# ]

print(midsv.transform(path_sam, qscore=True))
# [
#   {'QNAME': 'inversion',
#   'RNAME': 'example',
#   'MIDSV': '=A,=C,=G,=T,=A,=c,=g,=t,=A,=C',
#   'QSCORE': '15,16,17,18,19,20,21,22,23,24'}
# ]

🧩Helper functions

Read SAM file

midsv.io.read_sam(path_sam: str | Path) -> Iterator[list[str]]

midsv.io.read_sam reads a local SAM file into an iterator of string lists.

Read/Write JSON Line (JSONL)

midsv.io.write_jsonl(dicts: list[dict[str, str]], path_output: str | Path)

Since midsv.transform returns a list of dictionaries, midsv.io.write_jsonl outputs it to a file in JSONL format.

midsv.io.read_jsonl(path_input: str | Path) -> Iterator[dict[str, str]]

Conversely, midsv.io.read_jsonl reads JSONL as an iterator of dictionaries.

Reverse complement MIDSV

from midsv import formatter

midsv_tag = "=A,=A,-G,+T|+C|=A,=A,*AG,=C"
revcomp_tag = formatter.revcomp(midsv_tag)
print(revcomp_tag)
# =G,*TC,=T,=T,+G|+A|-C,=T,=T

midsv.formatter.revcomp returns the reverse complement of a MIDSV string. Insertions are reversed and complemented with their anchor moved to the new position, following the MIDSV specification.

Export VCF

from midsv import transform
from midsv.io import write_vcf

alignments = transform("examples/example_indels.sam", qscore=False)
write_vcf(alignments, "variants.vcf", large_sv_threshold=50)

midsv.io.write_vcf writes MIDSV output to VCF and supports insertion, deletion, substitution, large insertion, large deletion, and inversion. Insertions longer than large_sv_threshold are emitted as symbolic <INS>, large deletions (or =N padding) use <DEL>, and inversions use <INV>. The INFO field includes TYPE or SVTYPE, SVLEN, SEQ, and QNAME.