DeepMirCut In Silico Processivity and Cleavage Support Dataset across microRNA Orientation Variants
Description
This dataset provides the comprehensive, computational profiling of 1,917 human precursor microRNAs from miRBase v22.1, evaluating Drosha and Dicer cleavage processivity across four distinct transcript configurations: 1. Native (MIR): Canonical 5′ to 3′ orientation as expressed from the endogenous genome or sense cassette. 2. Reverse-Complement (RC): 5′ to 3′ reverse complementation representing the antisense cassette design. 3. Reverse Control (REV): Sequence polarity inversion (5′ to 3′ reversed) preserving nucleotide composition and base-pairing potentials without complementary strand pairing. 4. Complement Control (COMP): Watson-Crick base substitution preserving polarity and secondary structure pairing while altering primary sequence identity. Cleavage site predictions were computed using DeepMirCut, a convolutional and bi-directional LSTM neural network architecture trained on Drosha and Dicer cleavage datasets, capturing sequence, secondary structure, and terminal loop motifs.