GOSAE: Genome Ocean Sparse Autoencoder

This repository contains scripts and tools for training and evaluating a Sparse Autoencoder (SAE) approach to extract genomic features from biological sequence data.

Overview

GOSAE (Genome Ocean Sparse Autoencoder) is designed to:

1. Extract meaningful features from genomic sequences
2. Interpret functional significance of these features
3. Support the GenomeOcean platform for biological data analysis

Data Processing Pipeline

The repository includes scripts for:

• Removing duplicate sequences (scripts/remove_duplicates.py)
• Selecting representative sequences per species (scripts/select_one_per_species.py)
• Splitting data into training and validation sets (scripts/split_train_val.py)

Usage

Data Preparation

1. Remove duplicates from your FASTA files:

   python scripts/remove_duplicates.py input.fasta deduplicated.fasta
   

2. Select one sequence per species (optional):

   python scripts/select_one_per_species.py deduplicated.fasta representative.fasta
   

3. Split into training and validation sets:

   python scripts/split_train_val.py deduplicated.fasta train.fasta val.fasta --ratio 0.7
   

Training

[Training instructions to be added]

Evaluation

[Evaluation instructions to be added]

Requirements

• Python 3.6+
• BioPython
• [Other dependencies]

Citation

If you use this code in your research, please cite: [Citation information to be added]