Hewill Aligner

Hewill Aligner is a highly optimized Burrow-Wheeler Aligner (implement using modern C++20) specifically for paired-end short read alignment with read length ~150 bp.
The elapsed time of mapping sequencing data with 50× coverage (2×180GB) is less than 2 hours under 80 cores and ~18 GB memory usage and has comparable performance to bwa-mem2.

Compiler

GCC >= 10.2
Intel Threading Building Blocks (sudo apt install libtbb-dev)

Run

build executable

$ g++-10 main.cpp -o hewill -pthread -ltbb -std=c++20 -O3 biomodern/ssw.cpp -Wno-ignored-attributes

Command

Command:
         ./hewill index <hs37d5.fa>
         ./hewill align <hs37d5.fa> <in1.fq> <in2.fq> <sam_prefix> <sample_name> <read_group_id> [insert_mean] [insert_var] [thread_num]
Required Arguments:
       <hs37d5.fa>          Reference sequence hs37d5.fa file path
       <in1.fq>             Read1 FASTQ path
       <in2.fq>             Read2 FASTQ path
       <sam_prefix>         SAM file output prefix
       <sample_name>        SAM file SM tag value
       <read_group_id>      SAM file RGID tag value
Optional Arguments:
       [insert_mean]        Mean insert size of the paired-end data, default value is 550
       [insert_var]         Variance insert size of the paired-end data, default value is 150
       [thread_num]         Number of threads, default value is return value of std::thread::hardware_concurrency()

index

Only support for uncompressed hs37d5.fa. (download)
The suffix array sorting time is less than 3 minutes under 80 cores and ~20 GB memory usage.
```
$ ./hewill index /mnt/fa/hs37d5.fa
```
align
Only support for uncompressed fastq.

The following command will generate HG001.1.sam, HG001.2.sam, … HG001.X.sam and HG001.Y.sam in /mnt/sam/HG001 folder.

// <hs37d5.fa> <in1.fq> <in2.fq> <sam_prefix> <sample_name> <read_group_id> [insert_mean] [insert_var] [thread_num]
$ ./hewill align /mnt/fa/hs37d5.fa /mnt/fq/HG001.1.fq /mnt/fq/HG001.2.fq /mnt/sam/HG001 HG001 1 550 150 80

Performance

PrecisionFDA Truth Challenge benchmark versus bwa-mem2:

Important

This aligner is highly optimized on the following sequencing characteristic (other datasets are not recommended):

read length: 148~150bp

If you know the mean and variance of the insert size of the sequencing data, we highly recommend you pass it into the aligner.