Seq-SymRF

Seq-SymRF predicts potential associations between microRNAs (miRNAs) and diseases by integrating miRNA sequence features and clinical symptom information into a sequence- and symptom-based random forest model for association prediction.

Key Features:

• Algorithmic approach: Employs a sequence- and symptom-based random forest model that uses features extracted from miRNA sequences and clinical symptoms.
• Data integration: Integrates 917 miRNAs and 339 diseases using a binary association matrix (MDA.mat) derived from HMDD3.0.
• Negative sample construction: Constructs negative samples via a clustering method based on Euclidean distance calculations to obtain balanced representatives.
• Model validation: Trains and validates the model using fivefold cross-validation.
• Performance metrics: Reports accuracy 98.00%, specificity 99.43%, sensitivity 96.58%, precision 99.40%, Matthews correlation coefficient 0.9604, ROC AUC 0.9967, and PRC 0.9975.
• Case-study validation: Validated with case studies on leukemia, breast neoplasms, and hsa-mir-21 with literature and dbDEMC verification (e.g., 19/25 for leukemia; 20/25 for breast neoplasms).

Scientific Applications:

• Disease prevention and diagnosis: Supports identification of miRNAs linked to diseases to inform early detection and preventive strategies.
• Treatment development: Guides therapeutic intervention research by revealing miRNA–disease relationships and mechanisms.
• Drug discovery: Facilitates virtual screening for miRNA-associated targets and lead compound identification in drug research and development.

Methodology:

Runs the Seq-SymRF.m script to construct a random forest model; extracts features from miRNA sequences and clinical symptoms; constructs negative samples via Euclidean-distance clustering; trains the model with fivefold cross-validation; and evaluates performance using accuracy, specificity, sensitivity, precision, Matthews correlation coefficient, ROC AUC, and PRC.