Rockel Jason S, Sharma Divya, Espin-Garcia Osvaldo, Hueniken Katrina, Sandhu Amit, Pastrello Chiara, Sundararajan Kala, Potla Pratibha, Fine Noah, Lively Starlee S, Perry Kim, Mahomed Nizar N, Syed Khalid, Jurisica Igor, Perruccio Anthony V, Rampersaud Y Raja, Gandhi Rajiv, Kapoor Mohit
Division of Orthopaedics, Osteoarthritis Research Program, Schroeder Arthritis Institute, University Health Network, Toronto, ON, Canada; Krembil Research Institute, University Health Network, Toronto, ON, Canada.
Division of Orthopaedics, Osteoarthritis Research Program, Schroeder Arthritis Institute, University Health Network, Toronto, ON, Canada; Department of Biostatistics, Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada; Department of Mathematics and Statistics, York University, Toronto, ON, Canada.
Ann Rheum Dis. 2025 May;84(5):844-855. doi: 10.1016/j.ard.2025.01.012. Epub 2025 Feb 12.
Primary knee osteoarthritis (KOA) is a heterogeneous disease with clinical and molecular contributors. Biofluids contain microRNAs and metabolites that can be measured by omic technologies. Multimodal deep learning is adept at uncovering complex relationships within multidomain data. We developed a novel multimodal deep learning framework for clustering of multiomic data from 3 subject-matched biofluids to identify distinct KOA endotypes and classify 1-year post-total knee arthroplasty (TKA) pain/function responses.
In 414 patients with KOA, subject-matched plasma, synovial fluid, and urine were analysed using microRNA sequencing or metabolomics. Integrating 4 high-dimensional datasets comprising metabolites from plasma and microRNAs from plasma, synovial fluid, or urine, a multimodal deep learning variational autoencoder architecture with K-means clustering was employed. Features influencing cluster assignment were identified and pathway analyses conducted. An integrative machine learning framework combining 4 molecular domains and a clinical domain was then used to classify Western Ontario and McMaster Universities Arthritis Index (WOMAC) pain/function responses after TKA within each cluster.
Multimodal deep learning-based clustering of subjects across 4 domains yielded 3 distinct patient clusters. Feature signatures comprising microRNAs and metabolites across biofluids included 30, 16, and 24 features associated with clusters 1 to 3, respectively. Pathway analyses revealed distinct pathways associated with each cluster. Integration of 4 multiomic domains along with clinical data improved response classification performance, surpassing individual domain classifications alone.
We developed a multimodal deep learning-based clustering model capable of integrating complex multifluid, multiomic data to assist in uncovering biologically distinct patient endotypes and enhance outcome classifications to TKA surgery, which may aid in future precision medicine approaches.
原发性膝骨关节炎（KOA）是一种具有临床和分子影响因素的异质性疾病。生物流体中含有可通过组学技术测量的微小RNA和代谢物。多模态深度学习擅长揭示多领域数据中的复杂关系。我们开发了一种新颖的多模态深度学习框架，用于对来自3种受试者匹配生物流体的多组学数据进行聚类，以识别不同的KOA内型，并对全膝关节置换术（TKA）后1年的疼痛/功能反应进行分类。
在414例KOA患者中，使用微小RNA测序或代谢组学分析受试者匹配的血浆、滑液和尿液。整合4个高维数据集，包括来自血浆的代谢物以及来自血浆、滑液或尿液的微小RNA，采用具有K均值聚类的多模态深度学习变分自编码器架构。识别影响聚类分配的特征并进行通路分析。然后使用结合4个分子领域和1个临床领域的综合机器学习框架，对每个聚类内TKA后的西安大略和麦克马斯特大学骨关节炎指数（WOMAC）疼痛/功能反应进行分类。
基于多模态深度学习对4个领域的受试者进行聚类，产生了3个不同的患者聚类。跨生物流体的包含微小RNA和代谢物的特征签名分别包括与聚类1至3相关的30、16和24个特征。通路分析揭示了与每个聚类相关的不同通路。4个多组学领域与临床数据的整合提高了反应分类性能，超过了单独的单个领域分类。
我们开发了一种基于多模态深度学习的聚类模型，能够整合复杂的多流体、多组学数据，以帮助揭示生物学上不同的患者内型，并增强对TKA手术结果的分类，这可能有助于未来的精准医学方法。
