Making sense of tophat output "align_summary.txt"

[yangjr]

I run tophat2 with “-g 20" on some data. As a result, the "align_summary.txt" says

Reads:
Input: 92302015
Mapped: 79916185 (86.6% of input)
of these: 17726355 (22.2%) have multiple alignments (22103 have >20)
86.6% overall read alignment rate.

Then I run some manual checking on the numbers by using samtools on the output "accepted_hits.bam" and "unmapped.bam". I tried

samtools view unmapped.bam | wc -l

Result is 12385846, looks close enough, although not perfect (12385846 + 79916185 = 92302031).

Then I tried command (a)

samtools view -f 256 accepted_hits.bam | wc -l

Result is 15255480

I also tried command (b)

samtools view -f 256 accepted_hits.bam | cut -f1 | sort | uniq | wc -l

Result is 13013262

My understanding about tophat's behavior is that when there are multiple equally good hits, it will randomly pick one hit as "best", and all other hits will be assigned the 256 bit flag (secondary hits). So presumably, if the number 17726355 in "align_summary.txt" is the number of reads, then it should be similar to the result of (b). Or if 17726355 in "align_summary.txt" is the number of hits, it should be at least 2 times the result of (a). But none of my two guesses are correct.

So my question is, what exactly does the number 17726355 means.

Please let me know if I missed anything. Thanks all for any comments or suggestions!

jianrong

[11xinqi]

I run tophat2 with “-g 20" on some data. As a result, the "align_summary.txt" says


Then I run some manual checking on the numbers by using samtools on the output "accepted_hits.bam" and "unmapped.bam". I tried

samtools view unmapped.bam | wc -l

Result is 12385846, looks close enough, although not perfect (12385846 + 79916185 = 92302031).

Then I tried command (a)

samtools view -f 256 accepted_hits.bam | wc -l

Result is 15255480

I also tried command (b)

samtools view -f 256 accepted_hits.bam | cut -f1 | sort | uniq | wc -l

Result is 13013262

My understanding about tophat's behavior is that when there are multiple equally good hits, it will randomly pick one hit as "best", and all other hits will be assigned the 256 bit flag (secondary hits). So presumably, if the number 17726355 in "align_summary.txt" is the number of reads, then it should be similar to the result of (b). Or if 17726355 in "align_summary.txt" is the number of hits, it should be at least 2 times the result of (a). But none of my two guesses are correct.

So my question is, what exactly does the number 17726355 means.

Please let me know if I missed anything. Thanks all for any comments or suggestions!

jianrong

Did you get the answer?

[CPCantalapiedra]

in my case, que I run

samtools view accepted_hits.bam | cut -f1 | sort | uniq | wc -l

without the -f 256, the number of reads matches the number stated at align_summary.txt.

What doesn't make sense to me is the multiple mapping reads, since align_summary.txt number does not coincide with NH:i (>1) nor MAPQ counts.

[marika]

Hi Guys,
Is that the output for a paired end alignment?