Fig 1.
The flowchart of EDTMDA to predict miRNA-disease associations.
MiRNA/disease features extracted from integrated miRNA/disease similarity and known miRNA-disease associations were inputs of our training model. M DTs were obtained from M base learnings and the average of prediction scores from all DTs were calculated as final prediction results.
Fig 2.
The pseudocode of EDTMDA to predict miRNA-disease associations.
Fig 3.
Performance comparisons between EDTMDA and other 12 prediction models (HGIMDA, RLSMDA, HDMP, WBSMDA, RWRMDA, MCMDA, MIDP, PBMDA, MaxFlow, LRSSLMDA, MiRAI and MDHGI) in terms of ROC curve and AUC based on local and global LOOCV, respectively.
As a result, EDTMDA obtained AUCs of 0.9309 and 0.8524 in the global and local LOOCV, which exceed all of the above previous classical models.
Table 1.
AUC results between EDTMDA and other methods under 5-fold CV.
Table 2.
AUC results of EDTMDA between with dimensionality reduction and without dimensionality reduction under three cross validations.
Table 3.
AUC results between EDTMDA and RF under three cross validations.
Table 4.
EDTMDA was implemented to predict potential miRNAs related to Esophageal Neoplasms based on known associations in HMDD V2.0.
The top 50 predicted miRNAs were verified in dbDEMC and miR2Disease. The first column records top 1–25 related miRNAs and the third column records the top 26–50 related miRNAs.
Table 5.
EDTMDA was implemented to predict potential miRNAs related to Kidney Neoplasms based on known associations in HMDD V2.0.
The top 50 predicted miRNAs were verified in dbDEMC and miR2Disease. The first column records top 1–25 related miRNAs and the third column records the top 26–50 related miRNAs.
Table 6.
EDTMDA was implemented to predict potential miRNAs associated with Breast Neoplasms as a new disease by removing all known associations containing Breast Neoplasms in HMDD V2.0 database.
The top 50 predicted miRNAs were verified in dbDEMC, miR2Disease and HMDD V2.0. The first column records top 1–25 related miRNAs and the third column records the top 26–50 related miRNAs.
Table 7.
EDTMDA was implemented to predict potential miRNAs related to Carcinoma Hepatocellular based on known associations in HMDD V1.0 database.
The top 50 predicted miRNAs were verified in dbDEMC, miR2Disease and HMDD V2.0. The first column records top 1–25 related miRNAs and the third column records the top 26–50 related miRNAs.
Table 8.
The number of validated miRNAs among top 10 and top 50 predicted miRNAs in case studies between under true labels and under label randomization.