Say Hello!
Don’t be shy.


AI turns promising biomarkers into high value diagnostics

AI turns promising biomarkers into high value diagnostics

“Every research team and clinician we meet seems to have their own favourite biomarker or biomarker panel,” says Professor Graham Ball (pictured), founder and CSO of Intelligent OMICS Limited. “Our new Panel Optimisation Service turns these promising markers into valuable commercial assets.” Biomarkers are the building blocks of personalised medicine – the genes or proteins that mark the differences between us in our reaction to specific diseases. However it is not sufficient to find a single biomarker or group of markers that appear to differ between individuals in a population and then expect to be able to use the marker or panel as an accurate diagnostic. The molecular biomarker panel problem is clear: too many markers in the panel and the panel can’t be implemented on available devices, but too few markers and performance is compromised:
• Some panels include 20 or more biomarkers and are too unwieldy to apply;
• Others rely on a single, crucial biomarker which can easily be incorporated in an In Vitro Diagnostic test but whose performance is not good enough for wide adoption – in many circumstances, humans differ too much for a single biomarker to work well.

So now the Intelligent OMICS team, led by Professor Ball, has adapted the company’s proven Machine Learning platform to provide an efficient Panel Optimisation Service called POSitive: POSitive is fast and economic, optimising existing panels to be fit for purpose. A panel optimised through the POSitive service will typically include 3-8 biomarkers achieving over 90% specificity and 90% sensitivity – exceptional performance, but with the minimum necessary biomarkers. Output is presented first in the form of a research summary explaining the underlying systems biology and is then incorporated into a software model, delivered to the customer in a form appropriate to their application.

POSitive can be applied to optimise diagnostic, prognostic or predictive panels.
The POSitive Standard Operating Procedure includes: collation of data (using client or public datasets); AI analysis using machine learning to model the molecular biology; and development of software output.
The AI step relies on sophisticated non-linear machine learning, fitting over 50 million models per hour, using the company’s proprietary and patented neural-networks algorithm to prioritise biomarkers and model the pathway of the system. The service has been applied in many areas of oncology, infectious diseases, pulmonary diseases, CNS and others, working with UK and international companies and research institutes.

Intelligent OMICS works closely with a leading developer of lateral flow and rapid diagnostic technologies in human healthcare. Our partner engages major UK universities to analyse data but sought input from Intelligent OMICS in order to improve their latest respiratory panel. The Intelligent OMICS team refined the panel, reduced the number of biomarkers and significantly improved its performance. Our partner is now progressing the new asset through to commercialisation. The POSitive approach has been applied across multiple disease areas and further case study details are available related to projects with DSTL, Astra Zeneca, Syngenta, Public Health England and others.

The Intelligent OMICS approach uses patented Artificial Neural Network Algorithms. We incorporate architecture optimisation, rigorous regularisation, Monte-Carlo cross validation, early stopping, consensus and stability modelling and combined filter – wrapper methods to produce a model with optimised and generalised performance for blind validation data. Implication: low cost, high-value biomarkers, that work.

Contact: Dr Simon Haworth, Director, Intelligent OMICS
Email: [email protected]
Tel: +44 7802 183555
Nottingham: BioCity Nottingham, 1 Pennyfoot Street, Nottingham NG1 1GF
Wuhan, China: 康倍达(武汉)生物科技有限公, :武汉东湖高新技术开发区高新大道666号B1栋