Computational Biology

Our aim is to understand how individual bases in our genome predispose, alter and interact in normal and disease contexts. The computational frameworks developed in the group will drive the application of large public datasets in medical and translational research.

Advances in next generation sequencing (NGS) has lead to the routine generation of thousands of genome wide datasets for individual studies. Research groups focussing on a particular diseases rarely possess the analytical or computational capabilities to utilise these datasets in their work. Consequently vast amounts of publicly available data relevant to medical research is not being utilised to drive research and translation.

We are developing innovative data modelling techniques to facilitate data repurposing in translational studies. Highlights include:

In the FANTOM5 (functional annotation of the mammalian genome 5) project we used single molecule sequencing to transcriptionally profile a large number of primary cells, cell lines and tissues in human and mouse. We discovered a transcriptional signature that separates active enhancers from promoters and investigated the interplay between enhancer and promoter activation in a number of time-course experiments6. The expression atlas itself is the basis of a growing number of satellite publications.

We developed a first order pair hidden Markov model (HMM) based aligner to identify the correct mapping location of sequenced reads. The algorithm learns the probabilities of certain sequencing errors in a context specific manner from the raw data. Knowing which type of errors are made where allows for very accurate read mapping. Due to the use of machine learning techniques the algorithm can be automatically applied to thousands of datasets with varying sequencing quality without manual intervention.

The Encyclopedia of DNA Elements (ENCODE) project has systematically mapped regions of transcription, transcription factor association, chromatin structure and histone modification. We demonstrated that the vast majority of the genome is transcribed in different biological contexts. Protein coding genes express an average 10-12 isoforms at the same time and large classes of non-coding RNAs are specifically localised to sub-nuclear compartments.

Ongoing Collaborations (partial list):

• SeqNextGen: Translating NextGen Sequencing for the Diagnosis of Developmental Anomalies and Rare Diseases - our role is to develop better computational approaches to pinpoint causative variants.
• Single Cell transcriptomics - together with Jay Shin and Charles Plessy we are investigating the dynamics of novel non-coding RNAs during cell cycle using single cell transcriptomics.

We welcome targeted collaborations ideally involving large datasets allowing us to improve
the interpretation of data while strengthening our computational research.