Double Gyrase Analysis Pipeline

This repository contains scripts and data for analyzing double gyrase gene sequences and their mutations.

Project Structure

double_gyrae/
├── input/             # Input FASTA files and logits
├── figures/           # Generated plots and visualizations
├── scripts/           # Analysis scripts
├── output/            # Generated output files
├── run_evo.sh        # EVO model execution script
└── README.md

Data Generation Pipeline

1. Sequence Preparation

• Sequences are provided in input/double_genes.fasta with the following entries:
  • double_83_ref: Reference double gyrase sequence
  • double_83_TTT: Mutant sequence with TTT mutation
  • double_random: Random sequence of the same length

2. EVO Model Analysis

From the project root directory, run:

# Run the EVO model
./run_evo.sh double-genes input/double_genes.fasta input

This will:

• Create necessary directories under the specified output directory (e.g., input/output, input/jobs, input/figures)
• Submit the job to the EVO model
• Download results when complete
• Save logits as input_[sequence_name]_logits.npy inside <OUTPUT_DIR>/jobs/<JOB_ID>-v1/output/ (with the example above: input/jobs/double-genes-v1/output/)

For the R analysis scripts, ensure the .npy files exist at: input/input_[sequence_name]_logits.npy

If you used a different <OUTPUT_DIR>, copy or symlink the files from <OUTPUT_DIR>/jobs/<JOB_ID>-v1/output/ into input/ with the names input_[sequence_name]_logits.npy.

3. Data Analysis and Visualization

The analysis is performed using R scripts in the following order:

1. Initialize Data Processing

scripts/handle_EVO2_output_v1.R  # Handles logits file processing
scripts/evo2_analysis_functions_v1.R  # Core analysis functions

2. Main Analysis Script

scripts/EVO_learned_resist.R  # Generates all figures

This script performs:

• Loads and processes logits data
• Calculates log-likelihoods
• Generates difference plots between sequences
• Creates visualizations for:
  • Reference vs Random comparisons
  • Mutant vs Random comparisons
  • Second gene region analysis
  • Transition region analysis

Generated Figures

The analysis produces several PDF figures in the figures/ directory:

• double_ref-random.pdf: Comparison of reference to random sequence
• double_mut-random.pdf: Comparison of mutant to random sequence
• 83_TTT_2nd.pdf, 83_ref_2nd.pdf: Analysis of second gene region
• trans_*.pdf: Analysis of transition regions

Dependencies

• R packages:
  • Biostrings
  • ggplot2
  • reticulate (for Python integration)
• Python with numpy

Usage

1. Clone the repository
2. Place your FASTA files in the input/ directory
3. From the project root directory, run the EVO model:

./run_evo.sh double-genes input/double_genes.fasta input

4. Run the R analysis script:

source("scripts/EVO_learned_resist.R")

Notes

• This analysis of double genes indicates that evo2 7b memorizes the gene sequence within ~25bp and can then accurately predict every bp of the repeated gene going forward
• The highlighted nucleotides are the first 3 nucleotides of the repeated gene