Bioinformatics Team

MRC LMS

Thomas Carroll

The Bioinformatics Team.

  • Tom Carroll (ex-Head of Bioinformatics)
  • Gopuraja Dharmalingam (Epigenetics)
  • Sanjay Khadayate (Genes and Metabolism)
  • Yi-Fang Wang (Integrative Biology)
  • Marion Dore (Genomics)
  • TBD (Proteomics)

Websites

http://mrccsc.github.io/

Computing and Bioinformatics

http://bioinformatics.csc.mrc.ac.uk/

Where to find the team.

  • ICTEM
  • 2nd floor, MRC.
  • Central aisle,
  • Behind the printers.

Role

  • Experimental design.
  • Analysis
  • Bioinformatics Infrastructure.
  • Training.

Text

Experimental Design

“To consult the statistician after an experiment is finished is often merely to ask him to conduct a post mortem examination. He can perhaps say what the experiment died of.”

Fisher RA, 1938

  • Work closely with Genomics Team to help with design questions
    • Replicate number.
    • Sequencing depth.
    • Sequencing strategy.

Nice example experimental design

  • RNA-seq experiment (2014)
  • Graph shows major sources of variation.
  • Samples from same groups close together.
  • Samples from different experimental conditions separate. 

Nice example of experimental design

  • Smaller sources of variance relating to other metadata.
  • Samples group according to the day that RNA was extracted on.
  • Known effects can be removed from analysis.

Analysis

  • Initial data processing and QC.
  • Advice and support as needed.
  • Support throughout project.​

Increased demand for long term support.

Authorships in > 30 publication since 2014

Analysis support

  • Increased use of high throughput techniques in projects.
  • Greater use for bioinformatics in projects.
  • Analysis across project lifetime or individual elements.
  • Requires reproducible research.

Reproducible research

  • Reproducible results from computational methods should be straight forward.
  • Common problems.
    • Version and software changes.
    • Lack of analysis documentation.

rMarkdown

  • rMarkdown converted R (or python/many languages') code to dynamic reports.
  • Code, results and versions are reported within the same page.
  • HTML allows for inclusion of dynamic elements.

A do it yourself guide

Project tracking

  • Use Redmine software.
  • Multiple user interface to record project information.
  • Repository to version control scripts (SVN).
  • Wiki for internal documentation.

Infrastructure

  • Analysis pipelines.
  • Data delivery.
  • Software development.

ChIP-seq and RNA-seq

pipelines.

  • Common analysis steps can be automated.
  • Optimised for local resources.
  • Reproducible and comparable.
  • ChIP-seq and RNA-seq pipeline to automate alignment and quality control.​
  • Freely available for use or customisation on github

http://mrccsc.github.io/

Genomics Pipeline

  • Joined forces with Genomics to better automate some processes of basecalling/demultiplexing data. 
    • Basecalling - Convert optical data to sequence information.
    • Demultiplexing - Split sequence data by a sequence tag (index) which is unique per sample
  • Process is 
    • ​Users submit indexed samples and a sample sheet containing sample/index information.
    • Genomics provide a set of QCed sample sequence files for further analysis.

Why?

  • Reduce time consuming steps.
    • Sample sheet templates often misinterpreted by users or incorrectly filled out.
    • Thankfully not very originally and common errors can be automatically caught.
    • Manual creation of basemasks for index/read layout.
    • Remove dependence on Illumina's Windows' point and click software - Sequencing Analysis Viewer (SAV).
    • Reduce copy and paste.
  • Remove sources of potential error.
    • An automated process will be more consistent (in both accuracy and error).
    • Reduce copy and paste errors.

The MRC-LMS Genomics Pipeline

  • Automate samplesheet cleaning and updating.
  • Create basemasks or index/read description.
  • Parse results from basecalling/demultiplexing
    • Binary files from Real Time Analysis software.
    • XML files from BCL2fastq version 2.17 or above.
  • Summarise results to provide to user alongside resulting sequence files in fastq format.
  • Written in R, moved to Bioconductor in May 2017.

Others in the pipeline

  • Internal RNA-seq/ChIP-seq pipeline
    • Written in R.
    • Easily installed, maintained.
    • Allows Core to move between systems easily (both install and parallelisation).
    • Released soon.
  • Basecalling to ChIP/RNA-seq QC/initial analysis all in one reproducible, version controlled report.
  • Proteomics data pipeline?

UCSC genome browser

  • UCSC allows for visualisation of a range of genomics data types.
  • Public instances can be very slow.
  • CSC public instance maintained by Bioinformatics team.

  • web: http://ucsc

    FTP: ftp://ucsc

Software

  • Develop and maintain software relevant to our work.
  • R packages and javascript toolsets.
  • Release software to public (peer-reviewed) repositories.
    • Reviews improves code quality
    • Collaborative feedback.
    • Automated build reports and checking.

ChIPQC

  • Lack of suitable R/Bioconductor quality control tools for ChIP-seq.
  • Require methods to assess quality across high volumes of samples
  • ChIPQC developed and tested on 500 public datasets.

Package

Bioc2014 Tutorial

  • IGV is an popular alternate to UCSC.
  • Allows for inclusion of per sample metadata and complex sample display types.
  • Tracktables creates standalone and rMarkdown compliant tables.

Tracktables

Simple Example - Loading data

Simple Example - Loading data

Simple Example - Loading data

Simple Example - Loading data

  • Visualising genomics data over regions of the genome.
  • Allows for rapid generation of profiles and subsetting by IDs or other regions.
  • Arithmetic operations between and within profiles allows for rapid, iterative investigation of hypotheses.

Soggi

  • Peak calling in R is convenient.
  • Many peak callers in R have unwieldy input and far from optimised.
  • triform contains a reliable peak calling algorithm in need of optimisation for speed and long marks.
  • MRC CSC took over maintenance of triform in 2015

triform

basecallQC

  • Automate basecalling and demultiplexing.
  • Less manual intervention.
  • No dependence on proprietary software.
  • Customised for MRC-LMS use

Training

  • Aim to develop courses to meet MRC Clinical Sciences requirements.

    • ​R (8 courses)

    • Python (1 courses)

    • High throughput sequencing analysis. (3 courses)

  • ​Relevant and targeted training.

  • Real world examples.

  • Taught courses and material for online learning.

Material Repository

  • 2014 (R Basics)
    • Intro to R course.
    • Reproducible R.
  • 2015 (Basics of Bioinformatics)
    • ChIP-seq.
    • RNA-seq.
    • Bioconductor.
    • Genome Browsers and File Formats.
  • 2016 (Intermediate R and Bioinformatics)
    • Intermediate R.
    • Alignments and Counting.
    • Visualising High Throughput Sequeuncing Data
    • Introduction to Python

CSC Bioinformatics Course

  • Current and upcoming Bioinformatics training material can be found at our site

http://mrccsc.github.io/training.html

Github/Github.io/Travis/Appveyor and Training in 2017

  • All training is hosted on Github.
  • All pages automatically synced to Github.io
  • Training courses automatically tested on windows/mac/linux after any changes.

Training Collaborations

  • All material is open source (CC BY-NC-ND 4.0).
    • Free to distribute or collaborate.
    • No commercial use.
    • No remixes.
  • Bioinformatics Core Shared Training (with Mark Dunning)
    • Genomic File Formats
    • Intermediate R
    • ChIP-seq ->
    • IGV ->
    • -> Experimental Design

 

Training on the cloud.

  • Awarded grant from Amazon Web Services.
  • Use virtual linux servers to host  R and RStudio pre-loaded with course material.
  • Allow for larger, real world analysis tasks during training.
  • No need for dedicated classroom - train from anywhere.

Future!

  • Release and link RNA-seq and ChIP-seq pipelines.
    • Submit samples and get QCed, aligned data back.
  • Support proteomics and analysis of other MRC high throughput technologies.
  • Training updates
    • More python!
    • More non-programmer courses (e.g. Deeptools)!
    • Julia and other languages?!
    • Interactive online training!!
  • More analysis!!!

Have a great week!

Bioinformatics Team

Tom - thomas.carroll@imperial.ac.uk

Gopu - gopuraja.dharmalingam@imperial.ac.uk

Sanjay -  sanjay.khadayate@imperial.ac.uk

Yi-Fang - yifang.wang@imperial.ac.uk

Marion - marion.dore@imperial.ac.uk