Helomics, a subsidiary of Predictive Oncology, has begun sequencing tumor samples as part of its collaboration with UPMC Magee Women’s Hospital. The partners aim to use artificial intelligence (AI) to help better predict treatment responses for people with ovarian cancer.

The collaboration will use Helomics’s AI platform, which is called D-CHIP, to analyze genomic and treatment response profiles. Based on multi-faceted analysis of tumor data — for example, information on mutations, tumor gene expression, and tumor tissue configuration — the researchers are ultimately hoping to identify individuals who are more likely to respond to a given therapy. That would allow the most effective treatment to be given to each patient.

“These retrospective ovarian cancer cases were profiled Helomics as early as 2010; hence, we have 10 years’ worth of drug treatment data, survival and other outcome measures we are gathering from Magee’s clinical databases,” Mark A. Collins, PhD, the chief innovation officer at Helomics, said in a press release.

“We are now sequencing these cases, looking at both the tumor mutations (genome) as well as tumor gene expression (transcriptome) to build a comprehensive multi-omic picture of the tumor. We are also using deep learning on histopathology images of the tumor tissue (tissue-omics) to add an additional dimension to this multi-omic profile,” Collins added.

The hope is that the AI platform will be able to sift through all these layers of data in order to find patterns that are clinically meaningful. This, in turn, could help in the process of figuring out the most effective treatment option for each individual.

For instance, the AI may find that people with a certain combination of mutations in their tumors and laboratory test results tend to respond better or worse to a given therapy — suggesting that a person with the same set of characteristics is likely to respond in an equivalent manner.

Importantly, using artificial intelligence to do this analysis enables researchers to search for patterns that are simply too complicated for human identification, but are nonetheless useful.

“We believe the combination of the rich multi-omic profile of the tumor and clinical outcome data will allow us to build an AI-driven model of ovarian cancer capable of predicting the tumor drug response and patient outcome (prognosis),” Collins said.