Bioinformatics Seminars

Current Bioinformatics Seminar

20 November 2018

Clean Results in a world of small dirty data sets

Matthew Wakefield
WEHI Bioinformatics

Often in biology we are faced with imperfect data: ill defined groups, high variability measurements, confounded data and small sample sizes. As a case study I will present our recent publication on methylation predicting treatment response in cancer. This study illustrates how we used visualisation to maximise the information from a small scale but high confidence and unconfounded data set, generate a biological hypothesis and examine used the predictions made from this data to identify groups of patients that maximise the statistical power of the only available clinical trial data. Our approach confirmed our hypothesis explaining past failures to validate methylation of BRCA1 as a predictor of drug response.
These results are being applied in an ongoing clinical trial and I will describe the CRISPR capture Nanopore targeted sequencing approach we hope to apply to these samples, and how an AI 'sing song karaoke' scoring algorithm will be applied to this data.

Kondrashova et al 2018 Nature Communications :

Search past seminars