Medicine wants to move away from a `one-size-fits-all’ approach, and instead use the vast amounts of information now accessible to tailor treatments optimally: `the right drug at the right time for the right patient’ (precision medicine). To achieve this, just collecting data is not enough. One needs intelligent quantitative algorithms that can predict, from data of individual patients, how for each possible treatment regime their disease would evolve. Moreover, these algorithms must be interpretable.

 

The main barrier in extracting patterns from medical data is the need to see the difference between `signal’ and `noise’. When an algorithm mistakes noise for signal, this is called `overfitting’. The danger of overfitting happening increases if we include more data per patient or if the number of patients is not large enough. The conventional statistical methods used in medicine were not designed to deal with overfitting. 

​

Our approach to reliable inference from high-dimensional medical data is to develop new mathematical and statistical models, in collaboration with medical professionals and academia, designed to meet the challenges of modern data-driven personalized medicine. They combine the precision and interpretability of Bayesian approaches with mathematical tools from theoretical physics, and focus on:

 

• Undoing effects of overfitting in clinical outcome prediction for high-dimensional signals or small data sets. 

• How to decontaminate inferences for the effects of disease interactions.

• How to infer responder subgroups in clinical trials prospectively.

• How to identify optimal personalized treatments and drug doses.  

​

Below we give a more in-depth description of the technology behind our two main inference pipelines, spsSIGNATURE and spsMOSAICS.
Both have by now been used in multiple medical studies and publications.