Help understanding STAR algorithm

[sent...@gmail.com]

Hi ,

Please correct me, if I am wrong

First step of alignment is make a hg19 index from hg19 Fasta. i.e creating a "uncompressed Suffix Array (SA)" of the whole hg19 fasta ( human reference genome). 

Then for every read (or read-pair)  STAR looks for "exact-matches "/"matches" in the index and hence maps the sequencing read to the genome.

My question is : Every time STAR runs on a sample, it spends time to sorts the "uncompressed Suffix Array (SA) " ? I am wondering why doesn't STAR sorts the index when it builds it ? So that  for a new run it doesn't spend time sorting the SA and directly starts from the step  of matching the reads to the index ?

[Alexander Dobin]

Hi @sentisci

indeed, the suffix array is created only once, with the --runMode genomGenerate option.

On the mapping runs, the suffix array and other indexes are just loaded in RAM.

Cheers

Alex