MuRaL: inferring mutation rates via deep learning

MuRaL, short for Mutation Rate Learner, is a deep learning framework to predict mutation rates at the nucleotide level using only genomic sequences as input. MuRaL can build models with relatively few training mutations and a moderate number of sequenced individuals (e.g. ~100 individuals), and can leverage transfer learning to further reduce data and time demands. It can be applied to many species with population genomic data.

This documentation is for MuRaL v1.2.0.

Contents

• Overview
  • Framework Components
  • Output Demonstration
    • SNV Rate Prediction (A. thaliana)
    • INDEL Rate Prediction (H. sapiens)
  • Model Architecture
    • MuRaL-snv Architecture
    • MuRaL-indel Architecture
• Install
  • Using Conda
    • Prerequisites
    • System Requirements
    • Installation Process
  • Using Singularity
• Tools and examples
  • Command Line Tools
    • Main Commands
      • MuRaL-snv
      • MuRaL-indel
  • Model training
    • Input Requirements
      • Recurrent mutation BED format
    • Output data
    • Example 1: Basic Training
      • MuRaL-snv
      • MuaRaL-indel
    • Example 2: Training with large sequence
      • MuRaL-snv
      • MuRaL-indel
    • Example 3: Training acceleration
    • Example 4: Training with limited resources
      • MuRaL-snv
      • MuRaL-indel
    • Example 4b: Training with recurrent mutations
  • Model prediction
    • Input Requirements
    • Output data
    • Example 5: Mutation Rate Prediction
      • MuRaL-snv
      • MuRaL-indel
  • Transfer learning
    • Input Requirements
    • Output data
    • Example 6: transfer learning commands
      • MuRaL-snv
      • MuRaL-indel
  • Calculating k-mer and regional correlations for evaluation
    • K-mer correlation analysis
      • Input data
      • Output data
    • Example 7 : k-mer correlation commands
      • MuRaL-snv
      • MuRaL-indel
    • Regional correlation analysis
      • Input data
      • Output data
    • Example 8: regional correlation commands
      • MuRaL-snv
      • MuRaL-indel
  • Visualization of correlation results
  • Scaling MuRaL-predicted mutation rates to per base per generation rates
    • Example 9 : Scaling workflow
      • MuRaL-snv
        • Step 1: Calculate scaling factor
        • Step 2: Apply scaling
      • MuRaL-indel
        • Step 1: Calculate scaling factor
        • Step 2: Apply scaling
• Trained models and predicted mutation rate maps of multiple species
  • MuRaL-snv
  • MuRaL-indel
• Citation
• Contact