Abstract

Associations between microRNAs (miRNAs) and human diseases have been identified by increasing studies and discovering new ones is an ongoing process in medical laboratories. To improve experiment productivity, researchers computationally infer potential associations from biological data, selecting the most promising candidates for experimental verification. Predicting potential miRNA–disease association has become a research area of growing importance. This paper presents a model of Extreme Gradient Boosting Machine for MiRNA-Disease Association (EGBMMDA) prediction by integrating the miRNA functional similarity, the disease semantic similarity, and known miRNA–disease associations. The statistical measures, graph theoretical measures, and matrix factorization results for each miRNA-disease pair were calculated and used to form an informative feature vector. The vector for known associated pairs obtained from the HMDD v2.0 database was used to train a regression tree under the gradient boosting framework. EGBMMDA was the first decision tree learning-based model used for predicting miRNA–disease associations. Respectively, AUCs of 0.9123 and 0.8221 in global and local leave-one-out cross-validation proved the model’s reliable performance. Moreover, the 0.9048 ± 0.0012 AUC in fivefold cross-validation confirmed its stability. We carried out three different types of case studies of predicting potential miRNAs related to Colon Neoplasms, Lymphoma, Prostate Neoplasms, Breast Neoplasms, and Esophageal Neoplasms. The results indicated that, respectively, 98%, 90%, 98%, 100%, and 98% of the top 50 predictions for the five diseases were confirmed by experiments. Therefore, EGBMMDA appears to be a useful computational resource for miRNA–disease association prediction.

Highlights

  • Associations between microRNAs and human diseases have been identified by increasing studies and discovering new ones is an ongoing process in medical laboratories

  • Performance evaluation The performance of Extreme Gradient Boosting Machine for MiRNA-Disease Association (EGBMMDA) was evaluated by leave-one-out cross-validations (LOOCVs) and fivefold cross-validation on the known miRNA–disease association dataset retrieved from HMDD v2.0

  • We presented the computational model EGBMMDA under the hypothesis that functionally similar miRNAs are likely to be related to similar diseases

Read more

Summary

Introduction

Associations between microRNAs (miRNAs) and human diseases have been identified by increasing studies and discovering new ones is an ongoing process in medical laboratories. Predicting potential miRNA–disease association has become a research area of growing importance. EGBMMDA was the first decision tree learning-based model used for predicting miRNA–disease associations. The results indicated that, respectively, 98%, 90%, 98%, 100%, and 98% of the top 50 predictions for the five diseases were confirmed by experiments. RNAs of about 22 nucleotides in length found in a wide range of species, including viruses, plants, and animals[1,2,3]. Deficiencies or excesses in miRNA expressions are correlated to Their regulatory mechanism involves base-pairing to sites abnormal biological processes and human diseases[8].

Methods
Results
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call