Tu183 - Accelerating Target Discovery in Immunology Using Target Immune Engine

Travis Ahn-Horst – Precision Medicine and Computational Biology – Sanofi; Nadine Biesemann – Type 1/17 Immunology and Arthritis Cluster, I&I – Sanofi; Corneliu Bodea – Precision Medicine and Computational Biology – Sanofi; Clement Chatelain – Precision Medicine and Computational Biology – Sanofi; Emanuele de Rinaldis – Precision Medicine and Computational Biology – Sanofi; Giorgio Gaglia – Precision Medicine and Computational Biology – Sanofi; Johanna Harrison – Precision Medicine and Computational Biology – Sanofi; Shameer Khader – Precision Medicine and Computational Biology – Sanofi; Samuel lessard – Precision Medicine and Computational Biology – Sanofi; Dongyu Liu – Precision Medicine and Computational Biology – Sanofi; Hamid Mattoo – Precision Medicine and Computational Biology – Sanofi; Nima Nouri – Precision Medicine and Computational Biology – Sanofi; Franck Rapaport – Precision Medicine and Computational Biology – Sanofi; Leonardo Rodrigues – Precision Medicine and Computational Biology – Sanofi; Virginia Savova – Precision Medicine and Computational Biology – Sanofi; Heming Xing – Precision Medicine and Computational Biology – Sanofi; Lu Zhang – Precision Medicine and Computational Biology – Sanofi

Abstract Text: The exponential growth of biological data offers unprecedented opportunities for understanding and discovering new treatments for immune-mediated diseases, but these can only be realized with a strategy that can seamlessly analyze and integrate multiple data layers. To meet this challenge, we have developed the Target Immune Engine (TIE), a machine-learning-based framework for integrated analysis of several orthogonal data layers, and automated prioritization of novel immune drug targets.

TIE integrates for each gene of the human genome and across nine immune conditions (Asthma, Atopic Dermatitis, COPD, Crohn’s Disease, Psoriasis, Rheumatoid Arthritis, Systemic Lupus Erythematosus, Type 1 Diabetes, and Ulcerative Colitis), a large set of orthogonal data layers extracted from over 1000 studies, 500,000 samples, and 1400 features, totaling 135 million data points. These include genetic evidences from genome-wide associations and expression quantitative loci, single-cell and bulk transcriptomics data, proteomics data, along with functional information, druggability and gene-disease association from knowledge graph. Through combined use of machine-learning algorithms informed by retrospective clinical data and a rule-based weighting system, TIE implements automated target assessment scores, enabling discovery of novel target hypotheses that would be hidden through single-layer data analysis.

In this presentation, we will introduce the general data and methodological framework, along with its target predictive performance and a few examples of novel target candidates.