Machine Learning for Precision Medicine

Machine learning is a discipline within the field of artificial intelligence (AI) where algorithms are applied iteratively to a large data set in order to automate analytical model building. With each iteration, the algorithm “learns” from the data and its performance is improved.

In precision medicine, machine learning can be applied to data repositories too large for the human brain to parse. The patterns found in those large data repositories can help researchers to draw conclusions and predict an event.

Machine learning applications can help understand clinical variables and/or molecular properties to predict:

  • Disease onset, status, and relapse
  • Efficacy and safety profiles of a treatment
  • Other patient characteristics

BSSI has utilized machine learning approaches in many projects:

  • Drug response and drug synergy prediction using DNA or RNA data and identify patient subgroups with improved treatment effect
  • Prediction of disease diagnosis or relapse after treatment using biomarker data for developing potential companion diagnostics or other diagnostic tests
  • Support of clients in presenting machine learning and predictive modeling results to regulatory agencies

Using historical clinical trial data, BSSI builds predictive models to increase efficiency of clinical trials in different ways, including:

  • Clinical trial enrichment: Using baseline data to stratify patients into sub-groups based on each patient’s predicted personalized progression as measured by the primary endpoint. Enrolling patients who are predicted to progress steadily along a common path.
  • Clinical trial randomization:  Designing prediction-based stratified randomization simulations to ensure trial arms are balanced.
  • Clinical trial analysis: Predicting future primary outcomes to create virtual control arm. Comparing actual outcomes following intervention vs. created virtual control arm provides a measure to evaluate efficiency before investing in pivotal clinical trials.
  • Advance modeling of clinical trial: Validating diagnostic and prognostic value of new biomarkers, quantifying placebo effect, etc.