Google meeting

Google meeting-

MRC CSC objectives

• Platform for bioinformatics training

• Project to analyse and report QC metrics across all ChIP-seq data.

Training

• MRC asking for all PhD students to receive bioinformatics training.
  • Introduction to R
  • Statistical inference in Biology.
  • Applied bioinformatics

• Post docs and senior scientists wish to acquire skills.
  • R
  • High-throughput sequencing analysis.
  • Simple statistics

Training from Bioinformatics team

• General training in genomics data
• Visualisation of high throughput sequencing data
• Training in the analysis of high throughput sequencing data in R

Analysis of HTS in R

• All courses are presented as in RStudio.
• Rpres format used to allow display of material within RStudio itself.
• All material public and version controlled on github
• Topics include
  • Introduction to R
  • Reproducible R
  • Introduction to Bioconductor
  • Introduction to ChIP-seq and RNA-seq

Analysis of HTS in R - Issues

• Very popular course.
  • Requires large room.
  • Requires computer systems set-up in advance.
  • Any changes require update to all computers.
• Analysis of HTS is computationally expensive.
  • Most course use dummy data.
  • Use of subsetted data limits usefullness of course to real world data.
• Data for course must be downloaded to every machine.
  • Often incorrect selection of data leads to problems.

Analysis of HTS in R - Solution

• Rstudio server.
  • Bring low power laptop simply to login to server.
  • Environment already set-up and easy to update.
  • All computation run on server. Allows for examples using multicore processing and more memory.
  • Central/shared directory for data.

An Rstudio-server?

• Where to get an Rstudio server.
  • Invest in MRC CSC server?
    • Not so scalable
    • Not useful for most of the year
  • Use cloud platform? Tried in BioC 2014 and received grant for MRC CSC courses.

HTS pipelines

• Rapid and reproducible results from high thoughput sequencing data is essential for any modern bioinformatics core.
• To achieve this we use version controlled pipelines optimised for our local compute resources.

ChIP-seq pipeline

• Previous versions of ChIP-seq pipeline were developed at Cambridge University by Thomas Carroll.
• Widely used in CRUK, Sanger and Cambridge University and now adapted for MRC CSC in London.
• Many of the tools within pipeline have now been wrapped up in R/Bioconductor packages maintained by MRC CSC.
  • ChIPQC.
  • Triform.
  • soGGi.
  • tracktables.

Updated ChIP-seq pipeline

• First pipeline used multiple tool sets and so was hard to version control and install.
• An R centric ChIP-seq pipeline has now been developed within MRC Clinical Sciences Centre.
  • This new pipeline runs as a single R markdown script and generates HTML reports linked to IGV.
  • Easily installed dependencies from Bioconductor and CRAN repositories.

Testing the ChIP-seq pipeline

• To test the first version of the ChIP-seq pipeline we analysed 1400 datasets and investigated QC metrics and their relation to both each other and related processing steps.
• This provided us with essential knowledge of which QC flags to be used in controlling ChIP-seq data.

Testing the R-centric ChIP-seq pipeline

• More recently new forms or ChIP-seq have emerged (MNAse seq, ChIP-exo) and with new technologies the sequencing output has rapidly grown.

• A large scale reanalysis of ChIP-seq data is required to gain a better understanding of how metrics relate to new ChIP methods as well as the increased length/depth and complexity of sequencing.

Testing the R-centric ChIP-seq pipeline

• Re-run study with new tools and new pipeline on all available ChIP-seq, ChIP-exo, MNAse-seq data.

• Analysis of results will be reviewed within the Bioinformatics team and with Shamith Samarajiwa (MRC Cancer Unit) and Ines De Santiago (CRUK Cambridge) .

• The full pipeline and all metric results will be freely available and published within Github. Analysis of QC flags published in a peer reviewed journal.