VariantAlign is a bioinformatics tool developed in the Shi Lab at the Monash Biomedicine Discovery Institute for efficient assignment of sequencing read pairs to a reference sequence library.
Unlike conventional read aligners, VariantAlign is designed to identify the best-matching reference among thousands of highly similar, short sequences (typically a few hundred bases in length). It is particularly suited for applications where reads are assigned to a library of closely related sequence variants, such as slightly mutated transcript sequences derived from the same gene or transcript.
For each paired-end read, R1 is used as provided, while R2 is first reverse-complemented so that both reads are in the same orientation as the reference sequences. The two ends of the read pair are then compared against every reference sequence in the library.
During comparison, only base substitutions (mismatches) are allowed—insertions and deletions (indels) are not considered. Each end of the read pair is evaluated at all possible positions along a reference, including partial overlaps, and the position with the highest number of matching bases is chosen as its best alignment.
For a given reference sequence, a read pair is considered mappable to that reference only if at least one of the two ends has no more mismatched bases than the mismatch threshold (default: 3) at its best alignment position. If this condition is not met, that reference sequence is not considered for the read pair.
Among all reference sequences that pass this filter, VariantAlign calculates the total number of matched bases across both reads (using their respective best alignment positions). The read pair is assigned to the reference with the highest total match count. If two or more reference sequences achieve the same highest total match count, the read pair is reported as unmappable to avoid ambiguous assignments.
Binary releases of VariantAlign are available from the GitHub repository: https://github.com/ShiLab-Bioinformatics/VariantAlign. The program can be installed by simply decompressing the downloaded package.
VariantAlign currently supports x86-64 Linux systems and requires /bin/bash. If bash is located elsewhere on your system, update the first line in run_VariantAlign.sh to point to the correct path.
The only entry for running VariantAlign is run_VariantAlign.sh.
Usage: ./run_VariantAlign.sh <IN:file1.fastq.gz> <IN:file2.fastq.gz> <IN:library.csv> <OUT:output.xlsx>
The input files file1.fastq.gz and file2.fastq.gz must correspond to the first and second reads of each paired-end read, respectively. Reads must be generated using a stranded protocol (forward–reverse orientation).
Referece sequences are provided in input file library.csv. This file must contain exactly two columns: the reference sequence (first column) and the sequence name (second column). Do not include a header row or column titles. All values should be plain text without quotation marks. All the reference sequences must be longer than the reads in the two fastq.gz input.
The output spreadsheet is written in output.xlsx.
The following parameters can be adjusted by editing run_VariantAlign.sh:
THREADS: Number of CPU cores to use. Default:8.MAX_READ_LENGTH: Maximum allowed read length in the input data. All reads must be ≤ this length. Default:151.MAX_MISMATCH: Maximum number of mismatched bases permitted. At least one read in each pair must have a number of mismatches ≤ this threshold. Default:3.
Below is an example of library.csv.
ATCGGTCAATCGTAGCTAATCGGTCAATCGTAGCTAATCGGTCAATCGTAGCTA,seq001
TACGGTCAATCGTAGCTAATCGGTCAATCGTAGCTAATCGGTCAATCGTAGCTT,seq002
CCCGGTCAATCGTAGCTAATCGGTCAATCGTAGCTAATCGGTCAATCGTAGCTG,seq003
......