基于语义的疾病表型相似性

Disease Phenotype Semantic Similarity

已知一种药物可用于治疗某疾病,则该药物可能对与该疾病具有相似表型的其他疾病有疗效。因此,大规模地计算疾病表型相似性可辅助发现的疾病新的治疗方法。我们从OMIM下载了3742种疾病的表型信息,从Mesh词库下载13721个关联解剖学和疾病症状的注释词。我们将以上的Mesh词逐一在3742种疾病的表型信息文本中搜索,得到每种疾病涉及的Mesh词汇列表,进而基于语义分析的方法系统地计算了疾病表型的两两相似性矩阵。我们发现疾病关联生物通路最多的有肿瘤生物通路,胰岛素信号通路,肥大心肌病通路和细胞粘附通路等。随疾病对表型相似度的增加,其更涉及相同KEGG生物通路的概率亦增加,证明了本文方法的可靠性。疾病表型相似性可作为疾病在基因水平相似性的补充,有望为药物发现研究提供一条新途径。

If a drug can treat a specific disease, this drug can probably treat the diseases with similar phenotype. Therefore, large - scale computing similarity of the disease phenotype can help to find the new treatment. We downloaded 3742 diseases phenotype information from OMIM database, and 13721 Mesh words related to anatomy and disease symptoms from Mesh vocabulary thesaurus. Each Mesh word was searched in all of 3742 diseases phe- notype annotation text. Finally, we got a Mesh vocabulary list for every disease. Then disease phenotype pairwise similarity matrix was systematically calculated using semantic analysis approach. We found that most of the diseases associated biological pathways include tumor biological pathways, insulin signaling, hypertrophic cardiomyopathy pathways and cell adhesion pathway. The probability of involving same KEGG biological pathway increases with the disease phenotype similarity. This also shows our method is reliable. Disease phenotype similarity can be used as the supplement of disease genomic space similarity, and may have potential application value in drug discovery.