Despite proven improvements in patients' response, combination therapies are underemphasized in cancer pharmacology. Challenges leading to this discrepancy includ the vast number of possible combinations, ineffective screening methods, and common toxicity issues in clinical trials. Serinus Bio utilizes a human-centric approach to uncover molecular mechanisms underpinning tumor resistance to single-drug treatments. They employ deep learning primed by disease-specific protein maps to model information flow from genotype to phenotype and automate resistance mechanisms discovery. With their approach, they accurately predict patient responses and effective drug combinations from monotherapy data, ultimately maximizing treatment efficacy while minimizing toxicity. Their technology is safeguarded by patents and trade secrets.
What is the problem?
Administering multiple antineoplastic therapies in combination has proven effective in driving deeper, more durable responses in cancer patients. Combined therapies can delay and prevent the emergence of tumor resistance. Despite the clear benefit to patients, combination therapies are not traditionally the primary focus of pharmacology pipelines given their associated challenges. First, combination therapies present a vast search space. For example, there are ~1.5 billion possible pairwise drug-dose combinations across only FDA approved anti-cancer therapeutics. Second, high-throughput screening of 2D culture systems for synergistic combinations has yielded limited success in identifying clinically trasnlateble small molecule combinations, let alone immunotherapies. Third, unacceptable toxicity is particularly common in clinical trials of anti-cancer drug combinations. Due to these challenges, combination therapy development is typically performed empirically through expensive and risky clinical trials.
What is their solution?
Deploying a human-centric approach, Serinus Bio deconvolves the molecular mechanisms of tumor resistance to monotherapies that reproducibly stratify patient responses in the clinic. Then, they algorithmically identify companion compounds with synthetically lethal behavior to tumor-drug molecular behaviousrs. The identified compounds are priorities by their therapeutic index and complementary profiles to the drug of interest. They finally predict dose ranges that will maximize efficacy while minimizing toxicity. Their key to success is their deep learning approach that automates the discovery of molecular resistance mechanisms by leveraging cancer-specific protein interaction hierarchies to model the flow of information from genotype to phenotype. The power of approach is highlighted by their ability to accurately predict patient responses and to predict synergistic combinations in high-throughput combination drug screens using only monotherapy data.
