cause of unmapped short with STAR

Question

[lkianmehr]

Hello,

Does anyone know the reason for having such results, 47% are unmapped short? need to mention I have cut the adaptors with cutadapt and trimmed them. the average length of reads is between 50-94.

Read   Mapped   Mapped   Mapped   Mapped Unmapped Unmapped Unmapped Unmapped
                    M/hr      number   length   unique   length   MMrate    multi   multi+       MM    short    other
Oct 29 13:58:22      0.3      126401      178    46.6%    167.1     2.2%     6.0%     0.0%     0.0%    47.2%     0.2%

[michael.ante]

Please see this comment here.

Answer 1

[Devon Ryan]

You'll need to play around with things like --outFilterMatchNminOverLread and --outFilterScoreMinOverLread, which set the minimum fraction of a read/pair that has to be aligned. Sometimes you have issues like this if the library prep was weird and you get lots of chimeric fragments.

Answer 2

[swbarnes2]

When STAR says "too short" what that really means is "didn't align". The most likely cause of that depends on what you were trying to do.

Note that STAR is fairly good at ignoring wrong sequence at the ends of a read, so proper adapter trimming is not going to rescue huge number of reads.

[lkianmehr]

with adding these options -outFilterScoreMinOverLread 0 --outFilterMatchNminOverLread 0 --outFilterMatchNmin 0 , I got such results

Time    Speed        Read     Read   Mapped   Mapped   Mapped   Mapped Unmapped Unmapped Unmapped Unmapped
                    M/hr      number   length   unique   length   MMrate    multi   multi+       MM    short    other
Oct 31 17:09:21      0.7      118472      197    74.4%     89.4     3.0%    25.3%     0.1%     0.0%     0.0%     0.1%
Oct 31 17:10:26      7.8     1539700      197    74.4%     89.8     3.0%    25.4%     0.1%     0.0%     0.0%     0.1%

mapped unique changed from 46% to 74 %, now you mean 46% is more trustable, however, mapped unique increased?

[Devon Ryan]

If you have to clip the heck out of something to get it to align you have to start wondering "why?" and whether the results can be trusted given that. Maybe there's a good reason for the "why?" question, if so you can probably trust the results, but you'll need to think critically about this yourself.