The complexity of the model systems sharply increases along the drug discovery process, while only a limited number of experimental models are available that accurately reflect human disease. Conventional in vitro and in vivo models cannot capture disease behavior in real patients, and tools – like CRISPR – don’t act like actual drugs. Drug discovery is costly, time consuming, model systems have poor translation rates to patients and do not significantly reduce the risk of failure in the clinic. This makes it incredibly hard to translate preclinical hypotheses to the clinic and create targeted drugs that truly help. This is where Turbine is creating a difference.

Before running any wet lab experiments, Turbine computationally simulates tumor cell behavior in patients to understand the complex mechanisms driving the disease. Simulations can reveal the right modality and combination approach to treat even the most resistant cancers. Observing these in silico experiments our biologists and translational experts gain insight into the molecular context by which mono- and combination therapies can potentially lead to patient benefit.

Simulated Cells™ can be used to run the equivalent of any preclinical or clinical protocol, at computational speed and scale. Running millions of simulations before conducting the most promising wet experiments, Turbine’s platform guides every step from target ID to clinical Proof of Concept.

Guiding the R&D process with simulations can increase the chance of success in the clinic. Pioneering an approach that combines simulation with machine learning, Turbine maps and model how thousands of signaling proteins interact characterizing cellular level cancer behavior and response or resistance to treatment. The company’s platform enables the simulation of drug-like effects from compounds that may not exist yet, on cells potentially unavailable for lab-based testing, like those of high unmet need cancer patients.

This approach will potentially allow to predict not only what works in cells, mice and people but more importantly, why and how. Continuous iterations of simulations and proprietary in vitro and in vivo experiments confirm predictions and progress the Turbine pipeline while simultaneously improving the underlying Simulated Cell™. As all programs and partnerships run on the latest version of the in silico cell model, training benefits accumulate, leading to a constantly improving platform. Using results to both generate the initial idea and to guide its iterations, as the models improve, this leads to a more rational process to understand the underlying disease biology. The company’s benchmarks show that simulations prevent 2 out of 3 failed experiments in vitro and every 2nd failure in vivo as well.

With the availability of large preclinical datasets on cancer drug sensitivity and gene essentiality, computational biology models for predicting cancer sensitivity are gaining popularity. However, comparing these models proves to be a challenging task, as there are numerous published models and methods available, making it difficult to conduct meaningful comparisons without reproducing them on your own data.

Armed with the experience of benchmarking its own models at Turbine, the company publish the Turbine Benchmark Suite. This carefully composed benchmark set focuses on models’ ability to identify biologically applicable predictions. “While this benchmark set is not entirely foolproof and can potentially be overfit with sufficient attempts, we have made substantial efforts to ensure its resilience.”

As a company, Turbine prioritize results based on true holdout train/test splits. Unlike random splits, the team believe that cell-, gene-, and drug-exclusive splits offer more meaningful insights in real-life scenarios for predicting cancer sensitivity. Emphasizing these splits enables researchers to evaluate model performance in situations that closely resemble practical applications. Instead of solely identifying universally ineffective or harmful drugs across all cells, the Turbine team measure per target node performance. This approach requires passing predictors to demonstrate selectivity. In other words, a successful predictor must exhibit the ability to discern the specific contexts in which drugs are beneficial or detrimental for predicting cancer sensitivity. “Bias detection and mitigation: To identify biases orthogonal to the measured metrics, we employ a so-called Bias Detector. For instance, models driven by general drug sensitivity of cell lines may pass the selective performance threshold, but they can only identify trivial drug-cell line associations. Our Bias Detector framework helps identify such bias-driven models. By adhering to these principles, our aim is to provide a benchmark that facilitates fair and meaningful comparisons of computational biology models in predicting cancer sensitivity. We encourage researchers to develop robust and selective predictors that transcend the limitations of bias and demonstrate their utility in real-world scenarios.”