Eigengene develops proprietary AI, and uses it to analyze cancer whole genome sequencing (WGS) data, in order to develop (i) personalized prognostics and diagnostics, and (ii) novel therapeutics.

Our algorithms do what no others can, namely, simultaneously find both the similarities and dissimilarities across multiple and diverse high-dimensional datasets, known as tensors. This enables them to find what other methods miss. For example, in cancer genomic data, our algorithms were able to identify multi-hundred-thousand marker DNA signatures that predict a patient’s survival and response to treatment. Our algorithms can scale to petabyte-sized datasets, yet are sensitive enough to detect meaningful patterns in cohorts of as few as 250 patients, each with a genomic profile of billions of nucleotides. This enables us to tap the full wealth of information that sequencing the whole genome provides. Most other methods have difficulty finding patterns in such “skinny” datasets, in which the number of measured attributes per patient is far greater than the number of patients. However, they are the norm in cancer genomics.

We analyze DNA copy-number profiles, which can be measured far more reliably than the more usually used RNA expression. The fact that our signatures comprise hundreds of thousands of markers makes them robust to measurement errors. Our signatures predict novel drug targets that correlate with patients’ outcome. The discoveries that the algorithms make come directly from the genomic data and are not biased by the clinical information. The algorithms can use data from any and all platforms, including those that are already FDA approved and deployed in pathology labs.

#startups #AI #drugdiscovery