Date: Sunday, November 5, 2017
Session Type: ACR Poster Session A
Session Time: 9:00AM-11:00AM
Background/Purpose: High-throughput gene expression profiling of skin biopsies from patients with systemic sclerosis (SSc) has identified four “intrinsic” gene expression subsets conserved across multiple cohorts and tissues. These are the inflammatory, fibroproliferative, normal-like, and limited subsets. In order to classify patients in clinical trials or for diagnostic purposes, supervised methods that can assign a single sample to a molecular subset are required. Here, we introduce a novel machine learning classifier which is a robust predictor of intrinsic subset and test it on multiple independent patient cohorts.
Methods: Three independent SSc cohorts (Milano et al. 2008, Pendergrass et al. 2012, Hinchcliff et al. 2013) with gene expression data and intrinsic subset assignments were carefully curated and merged to create a training dataset covering a broad set of 297 skin biopsies representing 97 unique patients. Supervised machine learning algorithms were rigorously trained and evaluated using repeated three-fold cross-validation. We performed external validation using two independent SSc datasets: Chakravarty et al. 2015, which contains 16 samples/8 patients and Gordon et al. 2015, which contains 12 samples/6 patients. Additionally, we validated the classifier on a cohort of SSc patients with gene expression data independently generated by Assassi et al. 2015 (102 samples/97 patients). We used weighted gene co-expression network analysis and g:Profiler to identify and functionally characterize gene modules associated with the intrinsic subsets.
Results: Repeated cross-fold validation identified gene expression features using multinomial elastic net and incorporated them into the final model which achieved an average classification accuracy of 88%. All molecular subsets were classified with high average sensitivity and specificity, particularly inflammatory (83.3% sensitivity, 95.8% specificity) and fibroproliferative (89.7% sensitivity, 94.1% specificity). Through multiple rounds of external validation, the classifier maintained an accuracy ranging from 70% to 85%. In a re-analysis of gene expression data from Assassi et al. study, we identified subsets of patients that represent the canonical inflammatory, fibroproliferative, and normal-like subsets. The inflammatory subset showed upregulated gene modules enriched in biological processes such as inflammatory response, lymphocyte activation, and stress response. Similarly, gene modules enriched for cell cycle processes were increased in the fibroproliferative subset.
Conclusion: We have developed a highly accurate and reliable classifier for SSc molecular subsets for single samples trained and tested on diverse cohorts comprised of 427 skin biopsies from 208 independent patients. These analyses show that the intrinsic gene expression subsets are a common feature of SSc found across multiple internal and external validation cohorts. Machine learning methods provide a robust and accurate mechanism for stratifying intrinsic gene expression subsets and can be used to aid clinical decision-making and interpretation for SSc patients and in clinical trials.
To cite this abstract in AMA style:Franks J, Martyanov V, Cai G, Whitfield ML. Novel Machine Learning Classifier Accurately Predicts Intrinsic Molecular Subsets for Patients with Systemic Sclerosis [abstract]. Arthritis Rheumatol. 2017; 69 (suppl 10). https://acrabstracts.org/abstract/novel-machine-learning-classifier-accurately-predicts-intrinsic-molecular-subsets-for-patients-with-systemic-sclerosis/. Accessed October 28, 2021.
« Back to 2017 ACR/ARHP Annual Meeting
ACR Meeting Abstracts - https://acrabstracts.org/abstract/novel-machine-learning-classifier-accurately-predicts-intrinsic-molecular-subsets-for-patients-with-systemic-sclerosis/