Identifying and predicting amyotrophic lateral sclerosis clinical subgroups: a population-based machine-learning study.
Identifying and predicting amyotrophic lateral sclerosis clinical subgroups: a population-based machine-learning study.
复制标题
DOI：
10.1016/s2589-7500(21)00274-0
复制
发表时间：
2022-05
影响因子：
30.8
通讯作者：
Chio, Adriano
中科院分区：
文献类型：
--
作者：
Faghri, Faraz;Brunn, Fabian;Dadu, Anant;Zucchi, Elisabetta;Martinelli, Ilaria;Mazzini, Letizia;Vasta, Rosario;Canosa, Antonio;Moglia, Cristina;Calvo, Andrea;Nalls, Michael A.;Campbell, Roy H.;Mandrioli, Jessica;Traynor, Bryan J.;Chio, Adriano
肌萎缩侧索硬化症（ALS）是已知的重叠综合征的集合。更好地了解这种异质性和区分ALS亚型的能力将改善临床护理并增强我们对该疾病的了解。基于经验观察提出了各种分类系统，但尚不清楚它们在多大程度上反映了ALS人群的亚结构。我们假设机器学习技术可以识别ALS亚型的数量和性质。我们应用无监督（统一流形逼近和投影，UMAP）、半监督（神经网络UMAP）和监督（基于LightGBM的集成）模型对2,858名意大利ALS患者进行了基于人群的队列研究，这些患者有详细的表型数据。我们在1097名意大利ALS患者的独立人群队列中重复了我们的发现。我们发现，将基于UMAP的半监督机器学习应用于多层感知器神经网络的输出，可以产生ALS患者的最佳聚类。这些聚类大致对应于Chiò分类系统定义的6种临床亚型（球型、呼吸型、连枷臂型、经典型、锥体型和连枷腿型）。在复制队列中发现了相同的群集。相比之下，其他ALS分类方案，如El Escorial分类、米兰-都灵临床分期（MiToS）和King临床分期，并没有充分地标记聚类。集成学习识别出12个预测ALS临床亚型准确率较高的临床参数（曲线下面积= 0.982,95%置信区间= 0.980 ~ 0.983）。我们的数据驱动研究提供了对ALS人群亚结构的深入了解，并证明Chiò分类系统可以稳健地识别ALS亚型。我们提供了一个互动网站（https://share.streamlit.io/anant-dadu/machinelearningforals/main），以便研究人员可以根据少量临床参数预测ALS患者的临床亚型。国家老龄问题研究所和意大利卫生部。
Amyotrophic lateral sclerosis (ALS) is known to represent a collection of overlapping syndromes. A better understanding of this heterogeneity and the ability to distinguish ALS subtypes would improve clinical care and enhance our understanding of the disease. Various classification systems have been proposed based on empirical observations, but it is unclear to what extent they reflect ALS population substructure. We hypothesized that machine learning techniques could identify the number and nature of ALS subtypes. We applied unsupervised (Uniform Manifold Approximation and Projection, UMAP), semi-supervised (neural network-UMAP), and supervised (ensemble based on LightGBM) modeling to a population-based cohort of 2,858 Italian ALS patients for whom detailed phenotype data were available. We replicated our findings in an independent population-based cohort of 1,097 Italian ALS patients. We found that semi-supervised machine learning based on UMAP applied to the output of a multi-layered perceptron neural network produced the optimum clustering of the ALS patients. These clusters roughly corresponded to the six clinical subtypes defined by the Chiò classification system (bulbar, respiratory, flail arm, classical, pyramidal, and flail leg ALS). The same clusters were identified in the replication cohort. In contrast, other ALS classification schema, such as the El Escorial categories, Milano-Torino clinical Staging (MiToS), and King’s clinical stages, did not adequately label the clusters. Ensemble learning identified twelve clinical parameters that predicted ALS clinical subtype with high accuracy (area under the curve = 0·982, 95% confidence interval = 0·980–0·983). Our data-driven study provides insight into the ALS population’s substructure and demonstrates that the Chiò classification system robustly identifies ALS subtypes. We provide an interactive website (https://share.streamlit.io/anant-dadu/machinelearningforals/main) so that researchers can predict the clinical subtype of an ALS patient based on a small number of clinical parameters. National Institute on Aging and the Italian Ministry of Health.