Skip to content

Latest commit

 

History

21 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DNAencoder GPT 🧬

Comparison of Deep Learning Architectures for Genomic Sequence Classification

📋 Overview

This project compares state-of-the-art deep learning architectures for classifying genomic regulatory elements: promoters, enhancers, and introns.

🎯 Objective

Traditional techniques (CNN, RNN) have limitations in capturing long-range dependencies in genomic sequences. We evaluate modern foundation models to overcome these challenges.

🔬 Models Evaluated

🏆 Encoder-based Models

  • DNABERT-2 (117M parameters) - Best Performance
  • Nucleotide Transform (500M parameters)

⚡ Convolution-based Models

  • HyenaDNA (7M parameters)
  • Caduceus (1.9M parameters)

📊 Results

Model Accuracy Precision F1-score
DNABERT-2 🥇 88.3% 88.8% 88.1%
Nucleotide Transform 🥈 85.3% 85.7% 85.5%
HyenaDNA 🥉 83.9% 84.1% 84.0%
Caduceus 72.2% 80.4% 69.3%

🗂️ Dataset

  • Introns: ~190k sequences from GENECODE
  • Enhancers: ~2k sequences from ENdb 2.0
  • Promoters: ~2k sequences from EPD

All sequences normalized to 512 base pairs.

🛠️ Methodology

  • Fine-tuning of pre-trained models
  • Train/Val/Test split: 80/10/10
  • Batch size: 16
  • Optimizer: AdamW with weight decay
  • Early stopping on validation loss

🔑 Key Findings

DNABERT-2 achieves the best performance across all metrics
✅ Transformer architectures outperform convolution-based methods
✅ All models (except Caduceus) exceed 83% accuracy

🚀 Future Work

  • Stratified k-fold cross-validation
  • Larger datasets for promoters and enhancers
  • Evaluation of Evo2 model (7B/40B parameters)

👥 Authors

Contributor 1
Casali Cristian
Contributor 2
Flotta Aldo

University of Modena and Reggio Emilia (Unimore)


Star this repo if you find it useful!

About

No description, website, or topics provided.

Resources

Stars

1 star

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages