Dr M. Adam Ali
Adam Ali is a Health Education England fellow in Medical Education at MedShr, the world’s leading discussion platform for doctors. He also holds honorary posts as an Honorary Research Assistant at UCL Institute of Ophthalmology: Ocular Biology and Therapeutics, and an Honorary Research Fellow at Moorfields Eye Hospital.
Recent advances in big data and machine learning have led to promising developments in the field of rheumatology, including earlier diagnosis, more accurate predictions on likely response to therapy, and accelerated drug development pipelines.
Big data and computational algorithms are becoming more advanced, and recent developments in rheumatology have been particularly exciting. High throughput technologies and machine learning infrastructure mean increased accuracy in classification of disease phenotype, earlier diagnosis before the setting in of advanced severe disease, more accurate predictions on which drugs can work for which patient, and accelerated drug development pipelines.
Rheumatoid Arthritis (RA)
Determining the best medications for patients with Rheumatoid Arthritis (RA) can be challenging due to contraindications and drug interactions as well as heterogeneity in different rheumatic conditions. Recent advances in use of genetic biomarkers may be able to help narrow down the right medications early for patients. A recent analysis found that an unsupervised machine learning method called non-negative matrix factorization (NMF) identified several gene clusters in RA patients which are activated differentially and can be used to predict the response to infliximab with high accuracy.
Similarly, big data and machine learning have been critical in guiding and accelerating the drug development pipeline for new RA therapies such as JAK inhibitors, offering critical alternatives or adjuncts to disease-modifying anti-rheumatic drugs (DMARDs) such as methotrexate.
Juvenile Idiopathic Arthritis (JIA)
Juvenile Idiopathic Arthritis (JIA) is an inflammatory condition of the joints seen in children. One recent study showed that a machine learning approach called “sparse multilayer non-negative matrix factorization” could use clinical information to stratify patients and even predict the likely response to therapy and disease course. Using clinical information on the pattern of joint involvement, this study found that disease was most associated with resistance to treatment if patients had inflammation in joints in the fingers.
Big data and machine learning have been critical in guiding and accelerating the drug development pipeline for new rheumatoid arthritis therapies.
How to make machine learning a reality in Rheumatology
A recent review suggests that machine learning can make precision medicine a reality sooner if clinical data is made more accessible to research teams. As an example, the Dialogue on Reverse Engineering Assessment and Methods (DREAM) community crowdsourced an open competition in which teams were invited to submit machine learning approaches to treatments in RA.
By providing a clinical and molecular dataset, this enabled the research community to develop and trial machine learning approaches which have shown much promise.
Management of RA patients during the COVID-19 pandemic
Patients with rheumatic conditions are often on immunosuppressive medication and steroids, which has posed a new threat to this cohort in light of the COVID-19 pandemic. As in other specialties, there has been rescheduling of routine care and prioritisation of only the most urgent services, with significant redeployment of staff to acute settings. For the patients themselves, a shift towards telemedicine has helped minimise patients’ risk of exposure to COVID-19 while offering uninterrupted care.
Likewise, rheumatologists, GPs and other healthcare professionals have turned to platforms such as MedShr to discuss cases, learnings and emerging therapies. In one such discussion group on the platform, “Rheumatoid Arthritis Discussion Group”, thousands of doctors around the world have been discussing their approach to complex clinical scenarios and sharing knowledge of RA.
Sources/references: Pandit A, Radstake TRDJ. Machine learning in rheumatology approaches the clinic. Nat Rev Rheumatol. 2020 Feb;16(2):69-70. doi: 10.1038/s41584-019-0361-0. PMID: 31908355.