DeepMir

DeepMir is a software tool that uses deep learning, specifically convolutional neural networks (CNNs), to classify microRNAs (miRNAs) into their respective families. The tool addresses several key aspects of applying deep learning to miRNA classification:

1. Feature encoding: DeepMir investigates whether incorporating predicted secondary structure information in the input matrix improves classification accuracy compared to using only primary sequence information with one-hot encoding.

2. Handling class imbalance: As there are many more non-miRNA sequences than miRNAs, DeepMir uses a softmax output layer to distinguish between in-distribution (miRNA) and out-of-distribution (non-miRNA) samples rather than assigning a negative class to all non-miRNA sequences.

3. Classifying small miRNA families: The tool examines the ability of deep learning models to classify sequences from small miRNA families correctly.

Topic

Functional, regulatory and non-coding RNA;Machine learning;Gene transcripts;RNA-Seq

Detail

  • Operation: Database search;miRNA target prediction;miRNA expression analysis

  • Software interface: Command-line interface

  • Language: Python

  • License: Not stated

  • Cost: Free of charge

  • Version name: -

  • Credit: City University of Hong Kong.

  • Input: -

  • Output: -

  • Contact: Yanni Sun yannisun@cityu.edu.hk

  • Collection: -

  • Maturity: -

Publications

Download and documentation

< Back to DB search