Workflow of Proposed Lightweight Bag of Tricks based Neural Network (BoT-Net) For lncRNA-miRNA Interaction Prediction.
Process of Generating Different Overlapping and Non-Overlapping k-mers .
AU-ROC Produced by Proposed BoT-Net Methodology using Nucleotides Selected Solely from Start, End, Stard-End, and Entire Region of LncRNA sequences without slicing miRMA sequences.
Performance of Traditional Fixed-size Sequence Representation Generation Schemes based on Minimum (Min-Len-Truncation), Maximum (Max-Len-Copy-Padding), and Average (Average-Len-Truncation/Copy-Padding) length of Sequences in terms of 7 Different Evaluation Metrics.
Performance Produced by Proposed BoT-Net Methodology using Nucleotides Selected Solely from Start, End, and Stard-End. First Row Represents Accuracy and F1-score, Second Row Shows the Specificity and Sensitivity figures.
BoT-Net: A very lightweight bag of tricks based neural network leverages a novel idea of generating precise yet highly
informative residue distribution based sub-sequences and a precise neural network to surpass state-of-the-art lncRNA-miRNA interaction prediction
performance by a significant margin. BoT-Net web server allows researchers and practitioners to perform in-detail analysis of lncRNA and miRNA sequences,
train and optimize the neural network from scratch, utilize pre-trained model to perform inference on test lncRNA-miRNA sequence pairs on the go,
and download a variety of descriptive, prescriptive and predictive artifacts during the active session