antifp2 1.1.0

AntiFP2: A tool for prediction of Antifungal Proteins

AntiFP2

AntiFP2 is a Python toolkit for predicting antifungal proteins using multiple approaches:

• ESM2 (fine-tuned language model)
• Machine Learning (PAAC features)
• BLAST (sequence similarity adjustment)
• MERCI (motif adjustment)

It provides both one-by-one protein prediction and batch processing.

⚠️ Note: For genome or metagenome pipelines (requiring Prokka integration), please refer to the GitHub repository or the Docker version.

---

🚀 Features

• Fine-tuned ESM2 model for protein sequence prediction

• Hybrid ML-based predictions using PAAC features

• Post-prediction adjustments:

  • BLAST: Sequence similarity to known antifungal/negative examples
  • MERCI: Motif enrichment detection

• One-by-one and batch prediction modes

• Logging of rejected or invalid sequences

---

📦 Installation

Install via PyPI:

pip install antifp2

Requires Python ≥ 3.12.

Ensure BLAST+ and MERCI binaries are properly configured if using BLAST/MERCI adjustments.

---

📁 Project Structure

antifp2/
│
├── antifp2.py               # Master script
├── antifp2_esm2.py          # ESM2-only predictions
├── antifp2_esm2_hybrid.py   # ESM2 + BLAST + MERCI hybrid predictions
├── antifp2_ml_hybrid.py     # ML + BLAST + MERCI hybrid predictions
├── blast_db/                # Preformatted BLAST database
├── MERCI/                   # MERCI motif files
├── envfile                  # Config file for BLAST/MERCI paths
├── README.md
└── setup.py

---

🧭 Usage

All pipelines are launched using the master script:

python3 antifp2.py --method <method-name> [options]

Available Methods

Method Name	Description
esm2	ESM2-only predictions (protein sequences)
esm2-hybrid	ESM2 + BLAST + MERCI hybrid predictions
ml-hybrid	ML (PAAC) + BLAST + MERCI hybrid predictions

For genome/metagenome pipelines using Prokka, see GitHub or Docker version.

---

1️⃣ ESM2 Only — esm2

Predict antifungal proteins using the fine-tuned ESM2 model.

Arguments:

Flag	Description
--input	Input protein FASTA file
--output	Path to output CSV file
--threshold	Optional probability cutoff (default: 0.5)
--no-cleanup	Optional: keep intermediate files

Example:

python3 antifp2.py --method esm2 --input proteins.fasta --output esm2_predictions.csv

---

2️⃣ ESM2 + BLAST + MERCI — esm2-hybrid

Hybrid prediction combining ESM2, BLAST, and MERCI adjustments.

Arguments:

Flag	Description
--input	Input protein FASTA file
--outdir	Output directory
--threshold	Optional probability cutoff (default: 0.5)
--envfile	Path to envfile with BLAST/MERCI paths
--no-cleanup	Optional: keep intermediate files

Example:

python3 antifp2.py --method esm2-hybrid --input proteins.fasta --outdir esm2_hybrid_results

---

3️⃣ ML Hybrid — ml-hybrid

Predict antifungal proteins using ML (PAAC features) with BLAST and MERCI adjustments.

Arguments:

Flag	Description
--input	Input protein FASTA file
--outdir	Output directory
--threshold	Optional probability cutoff (default: 0.6)
--envfile	Path to envfile with BLAST/MERCI paths
--no-cleanup	Optional: keep intermediate files

Example:

python3 antifp2.py --method ml-hybrid --input proteins.fasta --outdir ml_hybrid_results

---

📂 Outputs

File / Directory	Description
*_predictions.csv	Main predictions (probabilities, labels)
*_antifp2.fasta	Predicted antifungal protein sequences
blast_out.csv, *.locate	Intermediate BLAST/MERCI results
rejected_log.txt	Invalid or rejected sequences log

---

💾 Model Files

ESM2 models are automatically loaded from the package or can be downloaded from Hugging Face.

---

✅ Citation

If you use AntiFP2 in your research, please cite it appropriately.

---

👨‍💻 Support

• GitHub: https://github.com/raghavagps/antifp2
• Email: [e-mail]

---