One of our main differentiators from other statistical consulting groups in the pharmaceutical industry is our expertise in applications of Machine Learning related to drug discovery. Our previous experience includes:

• Developing ML classifiers for non-invasive diagnostics using physiological/radiological signals and images, genomics and other omics data. With our expertise in design of confirmatory trials and adaptive designs, we specialize in construction of seamless development and validation of such diagnostic devices;
• Use of ML for biomarker discovery for applications in precision medicine. This means that our support to the sponsors can start at the discovery phase and lead to biomarker-guided population enrichment designs;
• We have also previously used ML models for calculating predictive probabilities of trial success for interim decision making for complex longitudinal and survival studies, for estimating counterfactual treatment effects from RCT data and for mining of large multi-source real world data.

Our approach to ML is focussed on the end goal of developing diagnostics and biomarkers that have real clinical utility and thus focus on statistical aspects such as robustness, validity, generalizability and reproducibility and not just accuracy. For this reason we often limit ourselves in extracting big data features (predictors) that have biological relevance and in using explainable models. Apart from our core team of statisticians and data scientists, we also work with experts in Bioinformatics and Genetics as part of our network.

R/Python Package Validation Capabilities →