Predicting miRNA-disease associations using an ensemble learning framework with resampling method

Abstract

Motivation: Accumulating evidences have indicated that microRNA (miRNA) plays a crucial role in the pathogenesis and progression of various complex diseases. Inferring disease-associated miRNAs is significant to explore the etiology, diagnosis and treatment of human diseases. As the biological experiments are time-consuming and labor-intensive, developing effective computational methods has become indispensable to identify associations between miRNAs and diseases. Results: We present an Ensemble learning framework with Resampling method for MiRNA-Disease Association (ERMDA) prediction to discover potential disease-related miRNAs. Firstly, the resampling strategy is proposed for building multiple different balanced training subsets to address the challenge of sample imbalance within the database. Then, ERMDA extracts miRNA and disease feature representations by integrating miRNA–miRNA similarities, disease–disease similarities and experimentally verified miRNA-disease association information. Next, the feature selection approach is applied to reduce the redundant information and increase the diversity among these subsets. Lastly, ERMDA constructs an individual learner on each subset to yield primitive outcomes, and the soft voting method is introduced for making the final decision based on the prediction results of individual learners. A series of experimental results demonstrates that ERMDA outperforms other state-of-the-art methods on both balanced and unbalanced testing sets. Besides, case studies conducted on the three human diseases further confirm the ERMDA’s prediction capability for identifying potential disease-related miRNAs. In conclusion, these experimental results demonstrate that our method can serve as an effective and reliable tool for researchers to explore the regulatory role of miRNAs in complex diseases.

A clustering-based sampling method for miRNA-disease association prediction

“…To prove the ability of the CSMDA to predict potential disease-associated miRNAs, we compared it with six state-of-the-art MDA prediction models, including ABMDA ( Zhao et al, 2019 ), ANMDA ( Chen et al, 2021 ), GAEMDA ( Li et al, 2021 ), GBDT-LR ( Zhou et al, 2020 ), IRFMDA ( Yao et al, 2019 ) and ERMDA ( Dai et al, 2022 ). First, the CSMDA and other MDA prediction models constructed negative sample set by their respective methods.…”

“…Since there may be no functional similarity between two miRNAs, we integrated the miRNA functional similarity and the GIPK similarity of miRNA and . Inspired by previous works ( Dai et al, 2022 ), the integrated miRNA similarity between and was defined as follows: …”

“…In this work, the 5,430 experimentally confirmed miRNA-disease associations were taken as positive samples and the 184,155 unverified miRNA-disease pairs as unlabeled samples. Most methods ( Yao et al, 2019 ; Zhao et al, 2019 ; Zhou et al, 2020 ; Chen et al, 2021 ; Li et al, 2021 ; Dai et al, 2022 ) of constructing negative sample set are to randomly select some unlabeled samples as negative samples, or apply k-means clustering on the unlabeled samples and sample negative examples from the resulted clusters. However, these methods may introduce potential positive samples into negative sample set and lead to the performance degradation of the trained model ( Chen et al, 2021 ).…”

“…In the CSMDA, each miRNA-disease feature vector has 878 dimensions, which may contain a large amount of noise and redundant information. Inspired by previous research ( Yao et al, 2019 ; Dai et al, 2022 ), we performed feature selection based on random forest variable importance score on each training subset. First, we trained a random forest model on each training subset and sorted all features by the variable importance scores which were generated by the random forest.…”

“…Chen et al proposed an anti-noise miRNA-disease association prediction algorithm (ANMDA) which applied the k-means algorithm to cluster the unlabeled samples and selected negative samples equally from each cluster to reduce the noise ( Chen et al, 2021 ). Dai et al presented a resampling-based ensemble framework (ERMDA) which constructed multiple balanced training subsets by resampling and obtained the final prediction result by soft voting strategy ( Dai et al, 2022 ). Liu et al proposed a new novel method via deep forest ensemble learning based on autoencoder (DFELMDA) to predict miRNA-disease associations ( Liu et al, 2022 ).…”