different result of STAR/STARlong when query RNA5S1.fa(121bp)

[Yugui Wang]

When I query RNA5S1.fa(121bp) against with GRCh38.p7 with STAR/STARlong, the result was different.

STAR can find all 16 prefect match(right),but STARlong find only 2 perfect match(wrong).
STAR/STARlong both used the same genome index created by STAR.

Command param:
STARlong --runThreadN 48 --outBAMcompression 0 --outSAMtype SAM --genomeSAindexNbases 15 --outSAMunmapped Within --genomeDir /usr/bio-ref/GRCh38.p7 --readFilesIn RNA5S1.fa --outFileNamePrefix ./ --twopassMode Basic --chimSegmentMin 25 --outSAMstrandField intronMotif --outSAMattributes NH HI AS nM NM --outFilterIntronMotifs RemoveNoncanonical --outFilterMultimapNmax 200


Of cause , STAR is better for 121bp(<650bp), but the input fasta file contain different size of seq,
so STARlong will be better for this case.


Best Regards
Wang Yugui

[Alexander Dobin]

Hi Wang,

with the latest release 2.5.2a, you can try to reduce --winReadCoverageRelativeMin parameter to, say 0.3 (or even lower) to allow for more multi-mapping alignments in STARlong.

Still, the STARlong algorithm is less exhaustive than normal STAR, an may not find all alignments. A solution could be to split your file into <300b and >300b reads and map them separately with STAR and STARlong.

Cheers

Alex

[Yugui Wang]

Hi Alex

--winReadCoverageRelativeMin 0.25/0.05 still only find 2 prefect match of all 16.
so we have to split the input file into <300b and >300b reads and map them separately with STAR and STARlong

[Alexander Dobin]

Hi Yugui,

could you please send me this read - I will try to understand the problem with STARlong.

Cheers

Alex