My thoughts and ideas
My thoughts and ideas
In order for htseq-count to use bam files directly it needs pysam. This can be installed with pip but that is not available by default.
Then, On AMI install and test htseq-count with bam files:
sudo apt-get install python-pip sudo pip install pysam
An alternative install procedure that has been tested and worked is as follows. The above procedure is preferred and should be tried first.
cd ~/bin/ wget https://pysam.googlecode.com/files/pysam-0.7.5.tar.gz tar -zxvf pysam-0.7.5.tar.gz cd pysam-0.7.5/ python setup.py build sudo python setup.py install
For R and other applications it would be nice if X11 worked. Note the install instructions for R would need to change as well.
Create a wiki section and exercise that summarizes read trimming concepts. Start with some raw data, including aligned reads. Align these reads without any trimming and assess alignment statistics using Picard, FastQC, etc. Now take these same reads and perform both adaptor trimming and quality trimming. Re-align the trimmed reads and assess the effect of trimming on alignment metrics.
Note: we do have an okay trimming section now, but it should really be starting with raw reads, rather than those that had already been proven to align to chr 22. This makes it harder to see the value of trimming by comparing alignment pre/post trimming. Instead we should perhaps get chr22 matching reads by k-mer analysis (kallisto), then trim (adapter and/or quality), then align trimmed and untrimmed, and summarize the difference.
We should add a section about batch effects. Both detecting the presence of batch effects as well as correcting for them during analysis.
Use the Snyder lab mouse/human tissue expression data as an example?
For convenience the cloud instances have been set up with very permissive security. Some better practices should be documented.
We previously had a fusion detection module but it was difficult to complete in time frames appropriate for a workshop. Further optimization is required. Another challenge is the lack of well engineered fusion detection software. This publication State-of-the-art fusion-finder algorithms sensitivity and specificity does a decent job of summarizing the current options available. Another caveat of this topic is that is mostly of interest to cancer researchers so it might only be included where there are sufficient students with this interest.
In particular we should add use of
Picard CollectRnaSeqMetrics (https://broadinstitute.github.io/picard/command-line-overview.html) and
RNA-SeQC (http://www.broadinstitute.org/cancer/cga/rna-seqc). It would also be good to include use of splicing metrics calculated from the TopHat junctions files. A standalone version of the TGI tool that does this would need to be created for this purpose.
Also use MultiQC to produce a combined report of the QC results: http://multiqc.info/
There a some nice slides/concepts that we could borrow from the BaseSpace Demo slides (see Obis ~/Dropbox Teaching/CSHL/2015/Workshop-CSHL-RNA-Seq-Metagenomics.pdf).
For example, pathway analysis of RNA-seq data, clustering, etc.
Gray lab breast cancer cell line dataset:
Introduction to RegTools functionality. After HISAT2 alignment, QoRTs (written in Scala) performs QC but also processes RNA-seq data to produce count files needed for splicing analysis by DEXSeq and JunctionSeq.