The benefits and pitfalls of machine learning for biomarker discovery.

Prospects for the discovery of robust and reproducible biomarkers have improved considerably with the development of sensitive omics platforms that can enable measurement of biological molecules at an unprecedented scale. With technical barriers to success lowering, the challenge is now moving into the analytical domain. Genome-wide discovery presents a problem of scale and multiple testing as standard statistical methods struggle to distinguish signal from noise in increasingly complex biological systems. Machine learning and AI methods are good at finding answers in large datasets, but they have a tendency to overfit solutions. It may be possible to find a local answer or mechanism in a specific patient sample or small group of samples, but this may not generalise to wider patient populations due to the high likelihood of false discovery. The rise of explainable AI offers to improve the opportunity for true discovery by providing explanations for predictions that can be explored mechanistically before proceeding to costly and time-consuming validation studies. This review aims to introduce some of the basic concepts of machine learning and AI for biomarker discovery with a focus on post hoc explanation of predictions. To illustrate this, we consider how explainable AI has already been used successfully, and we explore a case study that applies AI to biomarker discovery in rheumatoid arthritis, demonstrating the accessibility of tools for AI and machine learning. We use this to illustrate and discuss some of the potential challenges and solutions that may enable AI to critically interrogate disease and response mechanisms.

Interferon-α-mediated therapeutic resistance in early rheumatoid arthritis implicates epigenetic reprogramming.

OBJECTIVES: An interferon (IFN) gene signature (IGS) is present in approximately 50% of early, treatment naive rheumatoid arthritis (eRA) patients where it has been shown to negatively impact initial response to treatment. We wished to validate this effect and explore potential mechanisms of action. METHODS: In a multicentre inception cohort of eRA patients (n=191), we examined the whole blood IGS (MxA, IFI44L, OAS1, IFI6, ISG15) with reference to circulating IFN proteins, clinical outcomes and epigenetic influences on circulating CD19+ B and CD4+ T lymphocytes. RESULTS: We reproduced our previous findings demonstrating a raised baseline IGS. We additionally showed, for the first time, that the IGS in eRA reflects circulating IFN-α protein. Paired longitudinal analysis demonstrated a significant reduction between baseline and 6-month IGS and IFN-α levels (p<0.0001 for both). Despite this fall, a raised baseline IGS predicted worse 6-month clinical outcomes such as increased disease activity score (DAS-28, p=0.025) and lower likelihood of a good EULAR clinical response (p=0.034), which was independent of other conventional predictors of disease activity and clinical response. Molecular analysis of CD4+ T cells and CD19+ B cells demonstrated differentially methylated CPG sites and dysregulated expression of disease relevant genes, including PARP9, STAT1, and EPSTI1, associated with baseline IGS/IFNα levels. Differentially methylated CPG sites implicated altered transcription factor binding in B cells (GATA3, ETSI, NFATC2, EZH2) and T cells (p300, HIF1α). CONCLUSIONS: Our data suggest that, in eRA, IFN-α can cause a sustained, epigenetically mediated, pathogenic increase in lymphocyte activation and proliferation, and that the IGS is, therefore, a robust prognostic biomarker. Its persistent harmful effects provide a rationale for the initial therapeutic targeting of IFN-α in selected patients with eRA.

Diffusion enables integration of heterogeneous data and user-driven learning in a desktop knowledge-base.

Integrating reference datasets (e.g. from high-throughput experiments) with unstructured and manually-assembled information (e.g. notes or comments from individual researchers) has the potential to tailor bioinformatic analyses to specific needs and to lead to new insights. However, developing bespoke analysis pipelines from scratch is time-consuming, and general tools for exploring such heterogeneous data are not available. We argue that by treating all data as text, a knowledge-base can accommodate a range of bioinformatic data types and applications. We show that a database coupled to nearest-neighbor algorithms can address common tasks such as gene-set analysis as well as specific tasks such as ontology translation. We further show that a mathematical transformation motivated by diffusion can be effective for exploration across heterogeneous datasets. Diffusion enables the knowledge-base to begin with a sparse query, impute more features, and find matches that would otherwise remain hidden. This can be used, for example, to map multi-modal queries consisting of gene symbols and phenotypes to descriptions of diseases. Diffusion also enables user-driven learning: when the knowledge-base cannot provide satisfactory search results in the first instance, users can improve the results in real-time by adding domain-specific knowledge. User-driven learning has implications for data management, integration, and curation.

The successes and challenges of harmonising juvenile idiopathic arthritis (JIA) datasets to create a large-scale JIA data resource.

BACKGROUND: CLUSTER is a UK consortium focussed on precision medicine research in JIA/JIA-Uveitis. As part of this programme, a large-scale JIA data resource was created by harmonizing and pooling existing real-world studies. Here we present challenges and progress towards creation of this unique large JIA dataset. METHODS: Four real-world studies contributed data; two clinical datasets of JIA patients starting first-line methotrexate (MTX) or tumour necrosis factor inhibitors (TNFi) were created. Variables were selected based on a previously developed core dataset, and encrypted NHS numbers were used to identify children contributing similar data across multiple studies. RESULTS: Of 7013 records (from 5435 individuals), 2882 (1304 individuals) represented the same child across studies. The final datasets contain 2899 (MTX) and 2401 (TNFi) unique patients; 1018 are in both datasets. Missingness ranged from 10 to 60% and was not improved through harmonisation. CONCLUSIONS: Combining data across studies has achieved dataset sizes rarely seen in JIA, invaluable to progressing research. Losing variable specificity and missingness, and their impact on future analyses requires further consideration.