Go Back   SEQanswers > Bioinformatics > Bioinformatics

Similar Threads
Thread Thread Starter Forum Replies Last Post
Read counts from SAM file mapped to de novo assembled transcripts using HTSeq-count alan_sm RNA Sequencing 2 06-12-2015 08:54 PM
Strangely low proportion of mapped miRNA reads feralBiologist Bioinformatics 0 11-07-2013 04:05 PM
Discrepancy between HTSeq-count counts and total mapped reads M_staats Bioinformatics 1 03-21-2013 06:16 AM
htseq-count gets more reads? deepsea Bioinformatics 3 03-29-2012 11:27 AM
htseq count missing last mapped fragment mapardo RNA Sequencing 2 11-08-2011 05:27 AM

Thread Tools
Old 11-25-2013, 11:56 AM   #1
Location: UK

Join Date: May 2013
Posts: 10
Default HTSeq-count large proportion of mapped reads with no_feature

I am fairly new to RNA-Seq analysis and hoped someone could help me diagnose the reason why I have such a low proportion of reads mapped to features with my current method.

The experiment:
- Were looking at the retina in developing mice
- I have two conditions (wild-type vs mutant), each with 4 biological replicates (8 in total).
- RNA prepared using Illumina TruSeq stranded kit. Ribo-Zero Deplete techneque.
- Using a pair-end 2 x 100bp HiSeq reads. Approximately 40 million pairs per replicate.
The methods:
- Trimmed low quality regions/reads using Trimmomatic.
- Using TopHat2 to map reads to UCSC mm10 reference genome available on the tophat website.
  - --no-novel-juncs and larger --mate-inner-dist arguments
  - Example numbers from one replicate:
    - 92.8% overall mapping rate
    - 30,244,035 aligned pairs; of these:
      - 9.4% have multiple alignments
      - 1.9% are discordant alignments
    - 87.6% concordant pair alignment rate
    - These numbers seem pretty good to me.
- Sorted the acccepted_hits.bam file by read name using 'samtools sort -n'
- Converted bam to sam using 'samtools view'
- Extracting feature read counts using HTSeq-count and the gtf annotation file also from UCSC mm10.
  - Using default arguments (method=union and stranded=yes)
  - Example numbers:
    - successfull feature = 505,622
    - no feature = 29,039,281
    - ambiguous feature = 2525
    - non unique feature = 10,519,556
    - Success rate = 1.26%
As you can see I have only a very small proportion of reads mapped to a feature. I don't really know what is normal but this seems very small. I am certain that it is the correct genes.gtf annotation file as it was provided gzipped with the genome itself. What kind of % of reads mapped on features should I be expecting? What can I do to diagnose this problem? Can anyone spot anything wrong that I am doing? Thanks.

I also have a second question, I am able to do limited down-stream analysis using DESeq using the currently read counts. I get very few (only two) genes with differential expression. These two genes are directly adjacent to each other on the genome but not overlapping, these are Uckl1 and Znf512b ( There are 130 bp between them including the UTRs. I can't imagine how this may be happening if I am using the --union option in HTSeq-count as any reads that overlap both should be labeled as ambiguous features.

Thanks for any help.
edm1 is offline   Reply With Quote
Old 11-25-2013, 12:36 PM   #2
Senior Member
Location: USA, Midwest

Join Date: May 2008
Posts: 1,169

Here's your problem.
(method=union and stranded=yes)
For libraries prepared using TruSeq Stranded RNA Kits (dUTP second strand marking) you should set htseq-count -stranded=reverse
kmcarr is offline   Reply With Quote
Old 11-26-2013, 01:40 AM   #3
Location: UK

Join Date: May 2013
Posts: 10

Thanks. That has greatly improved the feature counts:

successfull feature = 15,000,966 (45.3%)
no_feature = 8,715,612 (26.33%)
ambiguous = 172,178 (0.52%)
alignment_not_unique = 9,215,369 (27.8%)

Does 45.3% success rate seem close to what I should be expecting in mice? I have nothing to compare it to.
edm1 is offline   Reply With Quote
Old 03-05-2014, 07:36 AM   #4
Junior Member
Location: Chicago

Join Date: Mar 2014
Posts: 3

Originally Posted by edm1 View Post
Does 45.3% success rate seem close to what I should be expecting in mice? I have nothing to compare it to.
Hey edm1,

I'm not sure if you were ever able to get advice on whether 45.3% is close to what you should expect, but here are my two cents:
1) The "alignment_not_unique" feature is for reads with low alignment quality, which usually means reads that had multiple mappings. Count-based methods of differential expression (e.g. DESeq) usually ignore these, i.e. they focus on "unique matches", so you shouldn't include them in your final percentage tally. Excluding them gives you a success rate of 62.8%.
2) I'm not very familiar with the literature on what an expected success rate would be, but one example I am aware of (and was posted in another forum) is this nature paper (link:, reporting in their supplementary information that they mapped an average of 86% (range of 64-91%) uniquely matching reads to exons. All of their numbers are reported in the first supplementary table. Thus, your success rate of 62.8% is close to the bottom of their range. There might still be some areas of improvement for you, but you're pretty darn close.

Hope that helps!
McG is offline   Reply With Quote
Old 05-26-2014, 08:28 AM   #5
Location: Ottawa, Canada

Join Date: Apr 2012
Posts: 22

Keep also in mind...this was Ribo Zero...All sorts of RNAs are in there...
I would expect PolyA selection to be higher (mapping to exons)...but that's just me thinking...

RemitoAmigo is offline   Reply With Quote

htseq-count, rna-seq

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

All times are GMT -8. The time now is 08:38 AM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2019, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO