"By concentrating on precision, one arrives at technique, but by concentrating on technique one does not arrive at precision” - Bruno Walter, German Composer
The Center for Precision Computational Systems Network (PreCSN) is radically changing how we sift through big data to find meaningful information. As the computational arm for the iNetMed’s drug discovery pipeline, PreCSN develops novel machine learning algorithms to drive precision drug discovery. PreCSN creates innovative computational approaches that use machine learning to hunt for disease targets, finding aberrant proteins and other molecules that traditional methods often miss. Also, these models can better predict whether a drug will succeed in phase III clinical trials.
To get there, researchers at PreCSN are creating disease maps. These maps are built using high-throughput data analysis techniques to better understand complex biological processes in normal and diseased tissues, identifying patterns in large datasets – separating signal from noise – and assessing them against fundamental principles underlying human biology. These analyses inform the creation of detailed maps of any disease and/or therapeutic area, highlighting universally conserved gene expression changes as tissues transition from healthy to pre-disease to overt disease. These maps enable researchers to visualize a condition from its earliest onset – years before it becomes symptomatic – identifying multiple, high-value therapeutic targets to halt and/or reserve disease progression. Picked out of universally conserved patterns, these targets are more likely to be relevant in most (if not all) patients in the clinic who are afflicted by that given disease.
While many labs investigate interactions between individual genes or clusters of genes, PreCSN does things differently. There are six possible fundamental relationships between gene clusters; while most look at two, PreCSN looks at all to capture more biological complexity. These relationships can illuminate disease targets, offering new ways for us to intervene/interrupt disease progression. They also provide the predictive horsepower to understand how diseases progress and assess a drug’s potential benefits, and how to strategize combination therapies or rescue therapies when current treatment is failing.
PreCSN has already made and is continuing to make fundamental discoveries that illuminate blood stem cell differentiation, bladder, colon, and prostate cancer, and other conditions. Some of these findings have been translated into diagnostic biomarkers to support early detection. In addition to target discovery, this approach can identify companion biomarkers and help select appropriate pre-clinical models to screen new drugs. PreCSN algorithms can rank models that best recapitulate human disease, providing superior tools to develop next-generation therapeutics.
Please visit our lab’s Inflammatory Bowel Disease (IBD) example, demonstrating our AI-assisted approach for target identification and validation.