Finding the right treatment early is vital for a positive prognosis. Concr’s work aims to speed up this process and ensure better outcomes for cancer patients.
Cambridge Enterprise Ventures portfolio company Concr has announced results from a new study published in Frontiers in Artificial Intelligence demonstrating that its platform, FarrSight®, more accurately predicts treatment responses for patients with pancreatic cancer than conventional biomarkers.
Concr is leading the adoption of AI digital twins to advance precision oncology in clinical research settings. By applying its proprietary Bayesian foundation model, FarrSight® generates omics-informed digital twins of individual patients by leveraging a multi-dimensional dataset of 25 billion data points across preclinical and clinical domains.
The new clinical validation results demonstrate that FarrSight® consistently outperformed conventional biomarkers, highlighting its potential to help oncologists accurately identify patients most likely to benefit from specific chemotherapies.
Founded by biomedical engineer Matt Foster, computational physicist Dr Matthew Griffiths and medical oncologist Dr Uzma Asghar, Concr uses established computational frameworks from astrophysics to interlink disparate and messy oncology data to allow scientists to confidently identify and develop biomarkers of drug response, with minimal data requirements.
Dr Irina Babina, CEO of Concr and coresponding author of the study said:
“We must remember that every single participant in a clinical trial represents a human life, not just a data point. By moving away from simply averaging out cohorts and instead using our digital twin model to simulate multiple therapeutic trajectories for the individual patient.”
"FarrSight® provides a far more accurate and comprehensive view than traditional population-based biomarkers. This opens a realistic path to true personalised medicine in both research and healthcare settings."
A shift in biomarker identification and application
The findings offer early evidence of a paradigm shift in how biomarkers are identified and applied. In an evaluation of patients with advanced pancreatic ductal adenocarcinoma (PDAC), an aggressive type of pancreatic cancer, Concr’s platform demonstrated a higher degree of accuracy in predicting outcomes for patients with the “basal-like” subtype, a population associated with poor survival rates and historically lacking effective predictive tools.
Based on data of patients enrolled in the COMPASS clinical trial, the study shows that Concr’s model achieved an area under the curve (AUC) of 72.3%, compared to a conventional biomarker AUC of just 44.8%, and an overall accuracy of 65.8% versus 47.4%.
Conventional biomarker discovery relies on a top-down approach, searching for broad, population-level similarities to determine who might respond to a drug. It also requires large sample size to draw such conclusions. In contrast, Concr’s platform (FarrSight®) employs a bottom-up methodology, focusing first on the complex, multi-omic interactions within an individual patient’s digital twin, and then scaling those highly individualised predictions up to stratify a wider cohort. This explains superior FarrSight® performance even in smallest datasets (study reports 38 patients with “basal-like” subtype).
Transparency in Concr’s AI model
The study also highlights the transparency and explainability of the AI. Rather than operating as a “black box”, FarrSight® breaks down feature importance across every data modality used in the analysis. This level of explainability provides researchers and clinicians with a clear, biological rationale for the model’s predictions, allowing them to see exactly which molecular features and gene expressions are driving every patient’s predicted response to therapy.
FarrSight® is currently available for research use only, helping biopharma partners to design smaller, more effectively powered clinical trials; discover novel biomarkers; and unlock rare cancer markets, while reducing development costs and accelerating the delivery of new therapeutics to market.
The work was conducted in collaboration with leading researchers from the Wallace McCain Centre for Pancreatic Cancer at Princess Margaret Cancer Centre, the Ontario Institute for Cancer Research, and the University of Glasgow.
