This article has been updated to reflect that Genomate's headquarters are in Cambridge, Massachusetts. This article has also been updated correct erroneous information provided by Genomate regarding CE marking of its computational model. The model is not a CE-marked medical device in the EU.
NEW YORK – Genomate is looking to partner with cancer centers and drugmakers in order to further prove the utility of its computational drug ranking tool.
The algorithm ranks targeted therapies that cancer patients are most likely to respond to based on their molecular tumor profile rather than a single biomarker.
It is currently standard practice in precision oncology to prioritize targeted treatments based on a specific tumor-driving alteration that has been shown to improve patients' outcomes. However, cancers can be driven by multiple genetic driver alterations, and the approach to ranking therapies using single tumor markers does not take into account the overall molecular profile of a patient's tumor.
Genomate's algorithm, dubbed digital drug assignment (DDA), ranks therapies that target driver alterations by calculating a cumulative score for each drug-profile match. In a paper published in NPJ Precision Oncology this month, the firm described how it calculates the aggregate score of multiple pieces of evidence linking molecularly targeted therapies to a patient's molecular tumor profile. Specifically, the score is calculated based on the number and the evidence level of the associations between the molecularly targeted therapies and all possible drivers and targets, and the aggregate evidence level of the associated potential driver genes and druggable targets in each patient.
According to Robert Doczi, Genomate's head of research, over the past seven years, multiple oncology practices at institutions in Central and Eastern Europe have incorporated DDA into their work, where molecular tumor boards are using it to make treatment recommendations for patients. Genomate is headquartered in Cambridge, Massachusetts with a satellite office in Budapest, Hungary.
As Doczi described it, the process starts when the lab returns next-generation sequencing results for a patient and the results are processed by DDA. The algorithm ranks the therapies and gives scores that delineate how well each drug is likely to work given the patient's molecular profile. The molecular tumor board then incorporates the DDA rankings into its consideration of which drugs to prescribe to the patient.
"It saved them a lot of time searching the literature because they had a very concise summary of what the literature-mining would reveal," specifically which drugs are associated with the molecular profile and what the strength of evidence is for or against each, Doczi said. The molecular tumor board can, of course, disagree with DDA's top-ranked drugs, but they often agree with the tool's rankings. "Obviously the clinical decision on the treatment is always in the hands of the treating physician," Doczi added.
Genomate isn't the only firm that has recognized the value of using computational tools to analyze patient data and guide cancer treatment. CureMatch, for example, uses machine learning and rules-based decision support to rank different combinations of drugs for patients based on their genomic test results and information from a proprietary curated database that includes an internal knowledgebase and input from public datasets.
However, Genomate CEO Istvan Petak distinguished DDA from other treatment matching tools by pointing out that while others are based on machine learning, DDA is not. Unlike a machine learning system, in which the decision process is often obscure, Petak characterized DDA as a transparent AI system that uses existing knowledge derived from published scientific data.
The company has validated DDA in multiple preclinical, clinical, and real-world datasets. While all these studies have provided important confirmation of its utility, Petak highlighted the company's reanalysis of data from the SHIVA01 clinical trial as a "singularity" in the history of precision medicine, because it suggested that a computational tool that analyzes genomic and outcomes data can help oncologists make better individualized treatment decisions for patients.
In the open-label, randomized SHIVA01 trial, sponsored by Institut Curie in Paris, researchers enrolled patients with metastatic solid tumors that were refractory to standard treatments and had at least one molecular alteration in the PI3K, mTOR, or RAF/MEK pathways. Patients were randomized to receive one of 11 molecularly targeted therapies matched to their genetic alteration or a physician's choice of therapy. The primary endpoint of the trial was progression-free survival.
The results of the trial were disappointing and are often cited by critics to argue that the utility of precision oncology is overhyped. There was just a few weeks' advantage in median progression-free survival for the group that received a molecularly matched agent in SHIVA01 compared to those that received their physician's choice of therapy, 2.3 months versus 2.0 months, respectively. The study investigators concluded that the use of targeted agents outside their indications did not offer an advantage in progression-free survival and that off-label use of those agents based on molecular profiling should be discouraged.
In June 2021, Petak and colleagues published a retrospective reanalysis of SHIVA01, using the DDA system to analyze the molecular and outcomes data of 113 patients in the study and score the molecularly targeted therapies they received. The authors underscored that SHIVA01 data were not used to train or optimize the DDA algorithm to avoid overfitting, which is a common problem with AI systems. They found that the DDA score was three times higher in patients who achieved disease control on the drugs they got in SHIVA01 than in those who progressed. The median progression-free survival was 3.95 months for patients with high DDA scores, compared to 1.95 months for patients with low scores. The findings were statistically significant.
Genomate researchers also published a study using real-world data collected from several pediatric oncology centers between 2017 and 2020. Samples from 103 patients were analyzed using whole-exome sequencing or targeted panel sequencing and analyzed by the DDA algorithm. A molecular tumor board evaluated the results and approved use of DDA-selected treatments in 56 out of 72 actionable cases.
"Now we have evidence that these tools can be incorporated into clinical practice," said Doczi, pointing out that a tool like DDA may also be useful in mitigating discordance in results returned by different molecular tumor boards when provided with the same molecular profiles. For example, in one study where molecular tumor boards from 12 leading cancer institutes were asked to make treatment recommendations for simulated cases, the concordance rate was just 62 percent.
With these and other clinical and preclinical validation studies in hand, Genomate is now looking to partner with cancer centers and pharmaceutical companies to apply DDA to patient care and clinical research. "What we can provide is an aggregated evidence level score of each treatment option that can [help] an oncologist choose between available [treatment] options," said Petak. "Oncologists would benefit … because there are more and more clinical situations where there are multiple treatment options that have not been compared to each other in randomized clinical trials."
Genomate has established a second headquarters in Boston and is eager to partner with oncology clinics, particularly community practices. "Community oncologists are very important because we can help them provide precision oncology close to the patient's home," said Petak. "We can also empower molecular tumor boards to expedite treatment decisions. Eventually, we want to democratize access to precision oncology for all cancer patients."
The company believes its tool also has promise in optimizing patient selection for clinical trials by streamlining the process of identifying eligible participants. It is interested in exploring collaboration opportunities with pharmaceutical companies for drug discovery and companion diagnostic development.
Genomate is working on regulatory strategies in both the US and EU.