Analytical Support to Predictive and Personalized Medicine
DNAlytics supports pharmaceutical companies in their drug design and validation by providing expert modeling, analytical and intensive computing (grid/cloud computing) know-how. This expertise mainly covers aspects of personalized medicine: population profiling (biomarker identification) and design and validation of models for automated prediction (diagnostics, prognostics, treatment response prediction, patients stratification, dose or adverse event prediction, ...). Our methodology can also be used to assess treatment efficacy. Our technology can be applied independently of the considered biological/pathological condition, and on a broad range of data types.
The main gains for our customers are an improvement of trial success likelihood, increased return on investment and a gain of time.
Clinical Research is undergoing a hard time. The expenses are dramatically increasing, while the return on investment is decreasing. The causes are the length of trials before approval with respect to the limited time of protection provided by patents, as well as the increasing risk of late stage failure (for safety and/or efficiency issues). DNAlytics proposes solutions to overcome that situation in most stages of clinical development. The following list proposes some examples in several development stages, but is not exhaustive:
- Enabling Translational Medicine: Our technology focuses on two objectives: identify biomarkers and building predictive models (for diagnostic, prognostic, response or adverse events prediction, etc.). Several potential applications in clinical or post-marketing situations are described below. However, DNAlytics technology can already be applied to really early research phases, on cell culture or animal models, and bring subsequent benefits for later phases (speedup, focus). Collaboration with academia and biotech companies to support evidence-based and translational medicine is thus totally relevant.
- Improving Early Stage Safety (Pre-clinical to Phase II): Based on observations on a set of individuals (cell culture, animal or human) among which some developed adverse events (AE) and others didn't, and on corresponding data (for example genomic), we can identify risk factors for developing adverse events, and at the same time build a prediction model for AE. The same approach applies to ADME. This model can be applied in subsequent phases in order to improve safety. The identified risk factors can be used to improve the considered drug/vaccine being designed.
- Increasing Proof of Efficacy Likelihood (Phase I and II): Typically, the efficacy of a treatment can be hard to prove in early clinical phases, given the limited number of individuals. It is specially true when dealing with complex or heterogeneous diseases (cancer, lupus,...). In that case, we can first identify markers discriminating between healthy volunteers and untreated diseased patients, for example based on high throughput gene expression data. After treatment, we focus on showing that treated patients are more similar to the healthy volunteers than they were prior to the treatment, based on the identified markers. At this stage, we can also identify responders and non-responders to the treatment, and identify robust markers that can be used for predicting this response (see the case below on predicting the immune response to a therapeutic vaccine). Models for AE prediction can here again be considered.
- Reducing Cost and Size in Late Stage Trials Drastically (Phase II and III): Based on markers for AE, response, diagnostic, etc. identified previously, the population can be stratified into several profiles (i.e. non-responders, responders with high risk of AE, responders with no such risk, etc.). Based on such stratification, several strategic choices can be made. Focusing on only some of the identified profiles will provide a high likelihood of success, on a subset of the population. In any case, being able to recruit patients corresponding to defined profiles will decrease the total number of patients to follow, and thus the time and cost of the trials.
- Enabling Innovative Strategies: Multi-Stage Launch and Companion Diagnostic (Phase III): The approaches described above allow considering new strategies for drug development and marketing. First, they can lead to much earlier market launch for some subsets of the population, allowing early return on investment, while trials continue on other profiles. They also allow co-development of a drug, and its companion diagnostic, for response or AE prediction.
- Improving Public Health Economics (Post-Marketing): The predictive medicine approach can be used to make decisions about the most adequate treatment among a given set of existing treatments for each patient with a given condition. Adopting that approach would benefit the patients by not giving them a treatment they would not respond to, or causing severe adverse event. This approach would also benefits public health organizations and health care insurers by conditioning reimbursement to the most appropriate treatment only. For PhIV, read our offer for Public Health.
- The scientific work on which is based this technology received IBM Belgium Computer Science Award.
- As an example, the application of our methodology led to a patent on responder/non-responder prediction in the context of a collaboration with GSK Biologicals on a therapeutic vaccine against melanoma.
- Read also the point of view of regulatory authorities on Predictive and Personalized Medicine (more precisely, on trials including Pharmacogenomics aspects), and the implications for reimbursement.