therincon / rexsearch

Port of vsearch into Rust

rexsearch

Port of vsearch into Rust

The licensing applies from vsearch, since it is mostly a port. I intend no copyright or other infringment. All credit to Torbjørn Rognes (torognes), also took some code from rust::bio. Same applies, all credit to Johannes. And of course, usearch.

Rewriting vsearch with protein comparison and FOGSAA. I have a few ideas how to use this for annotation and other projects. Such as "overhang" extension. I need to a tool to try to fill in gaps in assemblies from PacBio to Illumina. My Illumina reads always drop coverage in HiSeq and I would like to "fill-in" these gaps by returning partial alignments and long overhangs. Many tools already do something similar, but not quite the way I wish to use it though. Still getting to the point of the original intent. Now, I am also thinking about making several smaller, more modular tools, as part of package perhaps.

Comments

REXSEARCH is nearing VSEARCH speed. It is 30% faster for one core, but overall 50% slower because I only use 1 core. This is also for simple tasks such as msking or filtering reads, not alignment.

Candidate tools:

rfilter        fast filtering
rsearch     for sequnce matches
rsort          sort by length and abundance
rsample     subsample and shuffle reads
ralign         alignment and splicing tools

FILTER

Added capability to change softmask to hardmask (mask::fast_mask). Now have max/minlen filter (untested), max/minqual filter (untested), any_n filter (tested, filters sequences with any 'N' in seq), count_n (untested, LESS than provided, not equal to), and DUST preserving 80 char lines, and without (tested). DUST has soft- and hardmasking.

vsearch --fastx_mask /Users/daniel/Desktop/gg.fasta --fastaout  --qmask dust   10.55s user 0.13s system 291% cpu 3.660 total
./rsearch /Users/daniel/Desktop/gg.fasta  7.23s user 0.14s system 99% cpu 7.434 total

This in fact means vsearch was probably using more cores and was slower. Let's see with more thorough testing.

ALIGNMENT

Next Phase is back to alignment:

Smith-Waterman            Not Started
FOGSAA                          Not Started
Needlemann-Wunsch    In progress

CLUSTERING

This seems to be where vsearch and usearch shine. Not really going to focus on this for now.

DEREPLICATION

Now this could be something for rsearch to implement, rather simple, just look at "size=" tag on fasta header, then derep and eventually rerep. Saves space and time.

2 issues, vsearch --fastx_mask puts everything to uppercase (softmasking goes away). Maybe this is how it should be, but I don't think so. The other issue is, vsearch removes everything after " " in the fasta header. Again, maybe this is for a reason but I prefer to keep the file as is. Other than that, I am fairly sure I give more or less, the same output.