The future of medicine will be data-driven and personalised. Reaching this position efficiently and even for rare diseases where collecting vast datasets is all but impossible, however,  cannot be taken for granted. Extracting predictive patterns from medical data isn't easy: to have the best chance of effectively tailoring therapies to individual patients demands high-quality datasets and ambitious methods capable of extracting information from them optimally.

 

Numerous, disparate challenges are encountered by researchers who analyse clinical trial and epidemiological data, like distinguishing between signal and noise to avoiding overfitting, especially with high-dimensional datasets, latent heterogeneity and the influence of competing risks in a disease population. Traditional statistical methods and standard AI techniques often fall short when posed with these challenges, leading both to subpar data utilisation and unreliable reproducibility.  

 

Saddle Point Science develop cutting-edge mathematical and statistical models and algorithms to overcome these challenges. We work closely with medical professionals and academic researchers to understand the analytics problems they face, incorporating their real-world insights into our analytics software.

 

Our methods focus on overcoming overfitting, cleaning regression inferences of the effects of bias, identifying subgroups within disease populations, and determining treatment responders (those patients most likely to benefit from a particular treatment). Our main inference pipelines, spsSIGNATURE and spsMOSAICS, have proven their worth in multiple medical studies and publications, offering robust solutions for data-driven personalised medicine.