Signal Peptide Prediction using Machine Learning

Overview

This project focuses on the classification of amino acid sequences and/or the contained individual amino acids into various different categories related to signal peptide types and residue locations using various machine learning models.

---

Classification Tasks

• 2-state:

  • 0: Non signal peptide
  • 1: Signal peptide

• 4-state:

  • Non signal peptide
  • S: Sec/SPI
  • L: Sec/SPII
  • T: Tat/SPI and Tat/SPII

• 6-state:

  • I: Intracellular
  • M: Membrane
  • O: Extracellular
  • S: Sec/SPI
  • L: Sec/SPII
  • T: Tat/SPI and Tat/SPII

---

Model Versions

6-State Classifiers

---

4-State Classifiers

---

2-State Binary Classifiers

• simple one hot encoding combined with a linear layer