Questions on an agnostic AI system

A case study for an aortic aneurysm detection and other applications

Genuity Science is moving away from the old school methods of assessing disease states by individual tests through its artificial intelligence (AI) platform that draws on data from single cells to macroscopic phenotypes. By training AI and machine learning (ML) systems with disease‑agnostic methods, Genuity Science can expand the current understanding of biology from the cell to the phenotype level through use of Bayesian networks to perform probabilistic inference in high-dimensional biomedical data and uncover rules for molecular and whole-body governance. Furthermore, these AI/ML systems can generate hypotheses essentially free of confirmation bias, thereby allowing researchers in biology and medical fields to tease out novel mechanisms of action of diseases and therapeutic interventions.

In collaboration with investigators at Yale University Medical School, the Advanced AI Research Laboratory at Genuity Science has extensively studied cells which make up blood vessels and the causal gene drivers that lead to thoracic aortic aneurysms. In an experimental mouse-model system, results from the AI team’s time-series analysis of single-cell RNA sequencing (scRNA-seq) data showed that a single cluster of abnormal smooth muscle cells produced thickening of the vessel wall, which led to arteriosclerosis, and subsequently to thoracic aortic aneurysm.

Armed with this information, the team at Yale analyzed expression of specific protein markers by imaging mass cytometry to confirm the finding of a single cluster of abnormal smooth muscle cells in the aorta with consequent remnants of ossification and calcification. The Genuity Science AI team was then able to statistically construct, with the use of conditional probability, a causal gene dependency structure from the mRNA expression signature in that single cluster of cells that bulk RNA-seq methods would have otherwise missed. Their analysis uncovered a novel putative gene driver of disease etiology. Interestingly, a DNA variant within this gene associates with intimal hyperplasia of blood vessels, atherosclerosis, and hypertension in humans.

Along with cardiovascular disease, the Genuity Science AI team has also built integrated multi‑omic and single-cell ensemble AI/ML, digital pathology, biomedical imaging, and natural language processing strategies to identify etiological molecular mechanisms of cancer, COVID-19, nonalcoholic steatohepatitis (NASH), and neurodegenerative diseases. Capabilities in oncology applied to 8,200 tumors and 22 cancer types in the Cancer Genome Atlas (TCGA) discriminated any specific tumor from the other 21 types with greater than 99% accuracy. Moreover, the Genuity AI team reports an impressive 76% accuracy rate in predicting pan‑cancer patient survival at 60 months. Current state of the art technology is approximately 70% accuracy with patients of a single cancer type.

We asked Tom Chittenden and Jeff Gulcher to address a few questions on the Genuity Science AI platform and its applications in the biomedical sciences.

Read more.