Identifying ovarian cancer with machine learning RNA methylation pattern analysis

Proceedings in Obstetrics and Gynecology

article

Douglas J. Russo¹  Jesus Gonzalez Bosquet¹ 

[1] Dept. of Obstetrics and Gynecology, Carver College of Medicine, Univ. of Iowa Hospitals and Clinics

关键词: Machine learning;    ovarian cancer;    methylation;    informatics;   

DOI  :  10.17077/2154-4751.31885

学科分类：妇产科学

来源: University of Iowa

【 摘 要 】

High grade serous ovarian cancerremains one of the deadliest forms ofcancer among women, largely in partdue to the difficulty in early diagnosisand detection. Current methods fordetecting ovarian cancer rely onimperfect biomarkers with poorsensitivity and specificity. Newtechnologies are emerging which mayhave potential utility in improvingovarian cancer detection in the future.We hypothesize that we can modify anduse an existing Machine Learningframework to identify ovarian cancerbased on patterns found in themethylation of RNA extracted frombenign and malignant tissue samples.With access to a large biobank of samples we obtained RNA sequencesof 109 patients which were then used toperform unsupervised training of amachine learning framework with anAUC of 94.6 and interrogate theframework for top CpG sites influencingthe prediction using Shapley analysis.Prospective validation of this frameworkand investigation of top CpGs couldresult in a tool which could be applied toanalyze clinical samples for thepresence of ovarian cancer.

【 授权许可】

Unknown