Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Fusion-Driven Approach of Attention-Based CNN-BiLSTM for Protein Family Classification -- ProFamNet

Bahar Ali, Anwar Shah, Malik Niaz, Musadaq Mansoord, Sami Ullah, Muhammad Adnan

TL;DR

This study presents a model for classifying protein families using the fusion of 1D-CNN, BiLSTM, and an attention mechanism, which combines spatial feature extraction, long-term dependencies, and context-aware representations.

Abstract

Advanced automated AI techniques allow us to classify protein sequences and discern their biological families and functions. Conventional approaches for classifying these protein families often focus on extracting N-Gram features from the sequences while overlooking crucial motif information and the interplay between motifs and neighboring amino acids. Recently, convolutional neural networks have been applied to amino acid and motif data, even with a limited dataset of well-characterized proteins, resulting in improved performance. This study presents a model for classifying protein families using the fusion of 1D-CNN, BiLSTM, and an attention mechanism, which combines spatial feature extraction, long-term dependencies, and context-aware representations. The proposed model (ProFamNet) achieved superior model efficiency with 450,953 parameters and a compact size of 1.72 MB, outperforming the state-of-the-art model with 4,578,911 parameters and a size of 17.47 MB. Further, we achieved a higher F1 score (98.30% vs. 97.67%) with more instances (271,160 vs. 55,077) in fewer training epochs (25 vs. 30).

A Fusion-Driven Approach of Attention-Based CNN-BiLSTM for Protein Family Classification -- ProFamNet

TL;DR

This study presents a model for classifying protein families using the fusion of 1D-CNN, BiLSTM, and an attention mechanism, which combines spatial feature extraction, long-term dependencies, and context-aware representations.

Abstract

Advanced automated AI techniques allow us to classify protein sequences and discern their biological families and functions. Conventional approaches for classifying these protein families often focus on extracting N-Gram features from the sequences while overlooking crucial motif information and the interplay between motifs and neighboring amino acids. Recently, convolutional neural networks have been applied to amino acid and motif data, even with a limited dataset of well-characterized proteins, resulting in improved performance. This study presents a model for classifying protein families using the fusion of 1D-CNN, BiLSTM, and an attention mechanism, which combines spatial feature extraction, long-term dependencies, and context-aware representations. The proposed model (ProFamNet) achieved superior model efficiency with 450,953 parameters and a compact size of 1.72 MB, outperforming the state-of-the-art model with 4,578,911 parameters and a size of 17.47 MB. Further, we achieved a higher F1 score (98.30% vs. 97.67%) with more instances (271,160 vs. 55,077) in fewer training epochs (25 vs. 30).

Paper Structure

This paper contains 29 sections, 19 equations, 8 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related works and paradigms
  3. Traditional machine learning techniques
  4. The Naive Bayes (NB)
  5. The Radom Forest (RF)
  6. The Support Vector Machine (SVM)
  7. The K Nearest Neighbor (KNN)
  8. The Hidden Markov Model (HMM)
  9. Deep Learning Methods
  10. Deep Neural Network (DNN)
  11. The Recurrent Neural Network (RNN)
  12. The Convolutional Neural Network (CNN)
  13. The Long Short-Term Memory (LSTM)
  14. Bidirectional LSTMs (BiLSTM)
  15. Attention mechanism
  16. ...and 14 more sections

Figures (8)

  • Figure 1: Literature review
  • Figure 2: Visualizing ProFamNet: A graphical overview of layer structures
  • Figure 3: ProFamNet layer analysis: A structural overview
  • Figure 4: Embedding of partial protein sequence ...MDMMPFD...
  • Figure 5: Attention illustration by Bahdanau et al. bahdanau2014neural
  • ...and 3 more figures