Abstract

Critical to the success of the stratified medicine approach is a diagnostic programme that can reliably characterise molecules that act as markers for early disease detection and subsequent drug selection based on safety and efficacy. A number of large-scale facilities worldwide have been established to systematically discover these biomarkers, including the Stoller Biomarker Discovery Centre (SBDC) for proteomics in Manchester. Despite strict operating procedures that control sample preparation and analysis, analysis of biomedical proteomics and metabolomics data is challenged by the biological complexity, heterogeneity and dynamic range inherent in clinical samples, requiring large sample sizes for confident discovery, which in turn leads to issues with reproducibility. To overcome the limitations of existing informatics pipelines for robust identification, quantification and differential analysis, we have developed a novel workflow for biomarker discovery that for the first time extracts peaks and whole biochemical features through statistical signal processing of the unprocessed raw data. Recently, we have further extended this pipeline with a Bayesian modelling approach to assess peptide reproducibility for robust, reliable protein quantification and significance testing, and a new file format for scalable computation. In this talk I will discuss the translation of these methods into a platform for the SBDC , plus the potential of adapting these methods for corresponding applications in metabolomics.