Pharmaceutical research is increasingly generating extremely large data sets, especially as sponsors include pharmacogenomic and proteomic analysis in clinical protocols. Mining these rich data sets has the potential to unlock secrets that cannot be uncovered using traditional statistical techniques; giving new insights into genetic or gene expression signatures that can predict response and non-response, greater clarity on the drug’s mechanism of action, better safety and immune response profiles and generally getting a much deeper understanding of the biology of the situation.
Very large data sets also have a significant downside – data analysis is costly and time consuming, and there is great potential to get lost in endless work that leads nowhere. False positives abound, generating hypotheses that may require further work to confirm or deny. Outcomes need to be interpreted in ways that require new statistical approaches, with which sponsors and regulators may not be familiar, frustrating internal decision-making and regulatory approval.
Our philosophy is that the answer is in the data, provided the right questions have been asked and the right judgments made in whittling the data sets down to the manageable sizes. New bioinformatics techniques, including the use of a range of machine learning algorithms, can save significant time and effort in data analysis and provide incredible insight derived from patterns and relationships buried deep in the data. Our starting point is to think very carefully about the questions that need to be asked, and then design biomarker panels and other clinical data sets that yield actionable answers.
Bioinformatics and machine learning are extremely powerful tools. Synexa has the capability and experience to harness this power to help develop insights and advise on decisions about candidate therapies.