Chicken Epitope Platform
Deep learning for chicken immunopeptidomics

Predict chicken MHC Class I & Class II peptide binding

Upload a protein sequence or FASTA file and rank candidate peptides by predicted binding probability — built for vaccine developers, immunologists, and computational biology labs.

PyTorch-based · Computational Platform for Chicken MHC Binding Analysis

About the Platform

This platform hosts three independent prediction modules for chicken MHC peptide binding. BioAnchorFormer is a transformer-based deep learning model purpose-built for MHC Class I peptide binding prediction, combining sequence-level representation learning with an anchor-aware attention mechanism that emphasizes the residue positions known to dominate peptide–MHC binding affinity (typically P2 and the C-terminal anchor).

The MHC-II Predictor classifies candidate peptides into four classes — BLB1, BLB2, Shared, or Negative. Given any protein sequence, each single-protein module generates overlapping peptide windows, scores every candidate, and returns ranked results. Proteome Scan runs the same models across every protein in a multi-FASTA file at once, surfacing both the best individual peptides and the richest proteins — helping researchers prioritize epitopes for downstream validation in vaccine design and immunotherapy pipelines.

I & II
MHC classes
3 independent
Modules
Deep learning
Model type
PyTorch
Framework

Features

Everything a research lab needs to go from sequence to ranked binding predictions.

Flexible Input

Paste a raw protein sequence or upload a .fasta / .fa file directly.

Configurable Peptide Lengths

Choose from the peptide window sizes supported by each module.

Live Progress Tracking

Watch prediction progress, status, and estimated time remaining in real time.

Searchable Results

Instantly search, sort, and paginate through ranked peptide predictions.

One-Click Copy

Copy any peptide sequence to your clipboard for downstream tools.

CSV & Excel Export

Download full ranked results as CSV or Excel for offline analysis.

Dark & Light Mode

A clean, responsive interface that adapts to your working environment.

Validated & Secure

Strict input validation, file checks, and rate limiting protect every run.

How It Works

From raw sequence to ranked binding predictions in five simple steps.

01

Choose a Module

Pick MHC-I (BioAnchorFormer) or MHC-II prediction.

02

Provide a Sequence

Paste a protein sequence, upload a FASTA file, or paste a peptide list.

03

Choose Peptide Lengths

Select which peptide window sizes to generate and score.

04

The Model Scores Peptides

The module encodes each candidate peptide and predicts binding.

05

Review Ranked Results

Explore, search, visualize, and export high-confidence binding peptides.

Team

The BioAnchorFormer project is developed by a cross-disciplinary team spanning immunology, machine learning, and software engineering.

SD

Sarathkumar Devaraj

Researcher Student • UTA IMSE PhD Student

DSW

Dr. Shouyi Wang

Supervisor • UTA IMSE Faculty

DSNW

Dr. Stephen N. White

Collaborating Scientist • USDA

Contact

Questions about collaboration, model access, or the underlying research? Reach out.

Opens your email client addressed to sxd1910@mavs.uta.edu.