What if you could search hundreds of drugs to find personalized combinations tailored precisely to your patient? NCI-funded researchers are testing a new model, called BATCHIE (or “Bayesian Active Treatment Combination Hunting via Iterative Experimentation”), that could help you do just that.

BATCHIE not only enables you to discover how different classes of medications will interact, but it also helps you identify a host of promising candidates (and not just one).

According to the authors, the key to BATCHIE’s effectiveness is “active learning,” which means the model pools both known and unknown data points, then “actively” selects only relevant and informative new combinations to consider. This avoids the need to explore all possible combinations, which can be costly and time consuming.

When the researchers applied BATCHIE to existing data sets, the team found the model would have saved time by avoiding hundreds of thousands of experiments. In applying the model to pediatric sarcomas, the team explored the largest combination drug space to date and discovered potent combination therapies for Ewing sarcomas. But the developers think this is only the beginning for BATCHIE.

Senior author on the study, Dr. Wesley Tansey, of the Memorial Sloan Kettering Cancer Center, said, “Our model lets you explore medications much more efficiently. By predicting potential synergies and interactions, we can make smarter decisions in screening medications, potentially avoiding the need for extensive funding, detailed planning, advanced equipment, and years of experiments.”

He added, “BATCHIE’s performance in screening drug combinations was encouraging, but we think it has even greater potential. Combining BATCHIE with other models, such as those to help examine genomic and clinical features, could give us even greater screening efficiency—letting us better match patients who are most likely to respond to these combination therapies.”

NCI’s Division of Cancer Biology funded this work.