Contact us

Research

Pharmatics

Data analytics and machine learning

We are constantly improving our platform of machine learning methods, building on the latest advances in the field. We work on modifying the methods to achieve state-of-the-art performance even for small datasets. We are developing smart ways of using summaries and prior knowledge to improve performance, with a focus on prediction and knowledge discovery.

We are developing a new platform for evaluating causality in life-science studies, with the goal of speeding up identification of surrogate biomarkers and candidate drug targets. The platform will work with summary-level and patient-level data, and will complement the existing computational solutions for drug repositioning and repurposing.

Intelligent telemonitoring and telehealth

In partnership with the Telescot group from the University of Edinburgh and the Lothian Health Board, we have developed an algorithm to predict COPD hospital admissions using machine learning and telehealth data. Our model has halved the false detection rate of COPD admissions over the current symptom-counting scores. We have subsequently extended the model to include the Met Office weather data, and we will soon be adding new sources of telemonitoring information. Our machine learning approach adapts to personalized patient data and achieves a remarkable performance for predicting future COPD exacerbations. Our plaform can be used for developing other intelligent telehealth algorithms. See our IP Portfolio to learn more.

Data quality


We develop tools for improving the quality of clinical and biological data. In collaboration with Genos we have developed automatic peak detection and integration software that automates extraction of IgG glycans using Water's HPLC/UPLC liquid chromatography instruments. Our adaptive algorithm has increased the efficiency of signal extraction by several orders of magnitude, cutting the workload from several person-months to several machine-days for a typical cohort. We are extending the approach to other types of liquid chromatogrammes.

In collaboration with the University of Edinburgh, we have developed a high-performance tool for imputing extremely low-coverage sequencing data using genotype likelihoods. The approach does not require pre-phasing and achieves state-of-the-art performance using x10-x100 less time. The approach is based on a variational approximation from machine learning, and does not lead to an increase in memory complexity.

Drug discovery, drug development, molecular diagnostics


We are in multiple research and development collaborations on the applications of machine learning methods to biomarker discovery, prediction, and prognosis of chronic diseases. We are involved in projects on autoimmune diseases, colon cancer, oral cancer, diabetes, and metabolic health.

We have a joint patent application with the University of Dundee, Edinburgh, Cambridge, KCL, Pfizer, Roche, Sanofi, on a new-generation biomarker panel for complications of type 2 diabetes. The panel, identified by a machine learning approach, doubles the positive predictive value of identifying rapid progressors of kidney disease over an extended clinical model, and more than quadruples the precision over the population proportion.

In partnership with Erasmus Medical Center and Genos, we are preparing a patent application in the field of glycomics. Although we cannot disclose it yet, our invention may revolutionize glycomics and its use for detection and treatment of chronic diseases!

With the University of Glasgow and the University of Leeds, we are addressing complementary tasks of clinical decision support for autoimmune diseases. Additionally, with Glasgow, Edinburgh, and Stratified Medicine Scotland, we are analyzing next-generation sequencing data to understand the treatment response in rheumatoid arthritis.

With Edinburgh, Genos, and colleagues from the NHS Lothian, we are involved in a number of early-stage discovery projects aimed at identifying molecular predictors of clinical outcomes in cancers.

Health analytics


Working in a partnership with a group of health professionals, we are prototyping an AI-powered health analytics system that will help healthcare commissioners and board executives to manage the NHS.