Researchers have created a “digital twin” model constructed from the clinical and molecular profiles of patients with cancer that accurately predicted how a patient is likely to respond to a specific chemotherapy. The approach optimizes the treatment choice for patients using available clinical tests, according to the study authors. The study was presented during the 36th European Organisation for Research and Treatment of Cancer (EORTC)–National Cancer Institute (NCI)–American Association for Cancer Research (AACR) Symposium on Molecular Targets and Cancer Therapeutics (Abstract 11).

Study Methodology

The researchers leveraged technology called FarrSight^®-Twin, which uses astrophysics computational frameworks to model individual therapeutic responses to chemotherapy, integrating clinical data with a large gene panel or whole-exome and transcriptome sequencing. The researchers then created composite digital twins of clinical trial participants, including patients with early-stage and metastatic breast cancers, metastatic pancreatic cancer, or relapsed ovarian cancer, among others, to simulate eight historical clinical trials across various cancer types and treatment regimens.

The eight clinical trials used in the study were phase II or III prospective randomized trials with two study arms that compared different chemotherapy regimens, including anthracyclines, taxanes, platinum-based drugs, capecitabine, and hormone treatments. The technology team was unblinded to the study outcome in the first four studies and blinded in the last four simulations.

The researchers compared the predicted log odds ratio generated by the digital twin model for overall response rates (ORR) for each treatment arm against the reported log odds ratios from the actual trials. The researchers then applied the digital twin model to predict the optimal standard-of-care treatment for the different patient cohorts.

Key Results

The researchers found that the predicted log odds ratios for overall response rates in each treatment arm were concordant across eight clinical trials. Drug response rates showed significant improvement (ORR = 2.55, Fisher’s exact test P < .001), and overall survival increased with cohort enrichment (log rank P < .0001). The model accurately predicted overall survival across 23 solid tumors with an AUC of 0.78.

Further analysis demonstrated that patients in The Cancer Genome Atlas cohort who received treatments recommended by the digital twin model had significantly better therapeutic responses (Fisher’s exact test P < .0001) and survival outcomes (log rank P < .0001) compared with those receiving alternative standard-of-care therapies.

The study authors concluded: “The digital twin model can accurately predict treatment responses for different chemotherapy drug classes. This novel approach optimizes treatment choice for individual patients with current therapy using clinical tests that are currently available. In addition, digital twins can be used as synthetic controls in clinical trials, where the participant is [its] own control. This enhances interpretation of single-arm studies when a novel agent is combined with a [standard-of-care] backbone, facilitating combination studies of novel agents in earlier lines of therapy.”

Clinical Significance

Uzma Asghar, MBBS, PhD, MRCP

“We are excited to apply this type of technology by simulating clinical trials across different tumor types to predict patients’ response to different chemotherapies, and the results are encouraging,” said Uzma Asghar, MBBS, PhD, MRCP, Co-Founder and Chief Scientific Officer at Concr, manufacturer of FarrSight^®-Twin, and a consulting medical oncologist at The Royal Marsden NHS Foundation Trust, London.

“This technology means that researchers can simulate patient trials at a much earlier stage in drug development, and they can rerun the simulation multiple times to test out different scenarios and maximize the likelihood of success. It is already being used to simulate patients to act as controls for comparing the effect of a new treatment with the existing standard of care. We are currently developing this technology, so that it can predict treatment response for individual patients in the clinic and help doctors understand which chemotherapy will or will not be helpful, and this work is ongoing.”

Timothy Yap, MBBS, PhD, FRCP

In a statement commenting on the study results, Timothy Yap, MBBS, PhD, FRCP, a medical oncologist at The University of Texas MD Anderson Cancer Center, Houston, and Co-Chair of the EORTC-NCI-AACR Symposium, said: “Despite major improvements in cancer treatment, there are still many types of cancer where treatment options are limited. Designing and testing new cancer treatments is challenging, time-consuming, and costly. If we can exploit digital tools to make this process quicker and easier, that should help us find better treatments for patients more efficiently in the future.”

Disclosure: Funding for this study was provided by Concr Ltd, the manufacturer of FarrSight^®-Twin. For full disclosures of the study authors, visit www.aacr.org.