Which aligment tool is best for eukaryotic transcriptome?

Question

[sudha]

Dear one, Which alignment tool is best to align eukaryotic (homo sapiens) transcriptome? Please comment your answer with reason why! Thanks in advance!

p.s. My sequences are paired end & human samples! My intention is to do differential gene expression study!

p.s. I have seen that there are many alignment tools are used and given in publications, so confused a lot! I am new to bioinformatics!

[arta]

STAR is a good starter.

Answer 1

[Matteo Schiavinato]

Alignment of RNASeq reads on a genomic reference

• STAR (link) -> super-fast, gold standard for many people, awesome 2-pass mode (see manual)
• HISAT2 (link) -> same as Tophat2, much, much faster, good precision
• BWA (link) -> handles well also supplementary alignments (2048 bitwise flag) to then retrieve chimeras

all handle spliced alignment

Alignment of RNASeq reads on a transcriptomic reference

• RSEM (link) -> accurate quantification of gene and isoform expression (probably your best choice, imho, for this)

[lieven.sterck]

personally don't agree with the "all handle spliced alignment" statement , they all can handle split or gapped alignment but that's a totally different thing than spliced alignment (== splice-site aware aligner) !

[Matteo Schiavinato]

You are right, that is what I meant, my bad.

[sudha]

Thanks a lot for your answers, it made my day!

but (for alignment) RSEM uses bowtie2 only right ?

[Matteo Schiavinato]

Yeah, or bowtie 1, if you specify it. But it works quite fine, imho. However, you can use HISAT2 on transcriptome references tailoring the output directly for cufflinks (--dta-cufflinks) if this helps you.

Answer 2

[Carlo Yague]

You should also consider pseudo-alignments tools, such as kallisto and salmon. They are extremely fast and lightweight, can quantify millions of reads under a few minutes on a laptop computer.

Answer 3

[GenoMax]

BBMap suite (includes scanning/trimming program bbduk.sh and many things other than the aligner, which is bbmap.sh). I will not enumerate its advantages, since they should be apparent, if/when you try it.

Many aligners will need ~ 30G of RAM for human genome. If you don't have that available then bwa may be the one to try. It minimally needs ~6 G RAM.