Fastest (non-pseudo) aligner for RNA-seq illumina seq data (year 2019)

Question

[enxxx23]

What is the fastest (non-pseudo) aligner for RNA-seq illumina seq data today in year 2019?

[GenoMax]

fastest is a relative metric unless you do an apples-to-apples comparison on hardware you will eventually use. Choice of alignment options can have a significant impact on speed. It would be difficult to get those to align well among aligners.

bbmap.sh from BBMap suite will stand with the best of them on any given day.

bwa mem may be the smallest memory footprint aligner (~6-7G for human genome).

[enxxx23]

BWA MEM is not a RNA-seq aligner by design.

[GenoMax]

Sure. While the statement is true if you are looking for a splice-aware aligner it is not applicable in this case. Though you could use it if you were dealing with bacterial RNAseq data.

[enxxx23]

True, I should have mentioned the target organisms, which in this case are eukaryotes. Retroviruses do not have RNA so doing RNA-seq on retroviruses is the only choice! ;-)

[Nicolas Rosewick]

what do you mean by

Retroviruses do not have RNA

[WouterDeCoster]

I don't believe that speed only is a valuable concept. I could write an EXTREMELY fast aligner (with terrible accuracy).

[Nicolas Rosewick]

Challenge accepted !

[WouterDeCoster]

for read in SeqIO.parse("reads.fastq.gz"):
    pass

[Nicolas Rosewick]

ok great. Now submit to Nature Methods

[Kevin Blighe]

They'd accept it, for sure.

[Bastien Hervé]

At last try to :

read.seq == reference

[ddeemer]

if read in reference: print('Aligned!')

[Bastien Hervé]

STAR is for me, the best compromise between running time and efficiency

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5792058/

[i.sudbery]

I think STAR and HiSAT are probably similar. STAR requires less tuning, but HiSAT can be tuned to give very similar accuracy performance. The benefit of HiSAT is that it uses much less memory.

Answer 1

[Kristoffer Vitting-Seerup]

The better choices are:

1. Salmon
2. Kallisto

Both are very fast and highly accurate. Unless you have large differences in GC content there is no better tool - then they perform very similar. If you have a high GC-content Salmon is probably the better option due to its gcBias algorithm.

[WouterDeCoster]

OP explicitly mentioned being looking for aligners, not pseudo-aligners...

[Kristoffer Vitting-Seerup]

Woops - Missed the -non :D