DNA sequence classification by Deep Neural Network

Overview

DNA sequence classification by Deep Neural Network: Project Overview

  • worked on the DNA sequence classification problem where the input is the DNA sequence and the output class states whether a certain histone protein is present on the sequence or not.
  • used one of the datasets from 12 different datasets that we have collected. The name of the dataset is H3K4me2
  • To represent a sequence, we have utilized k-mer representation
  • For the sequence embedding we have used one-hot encoding
  • Different word embedding models: Word2Vec, BERT, Keras Embedding layer, Bi-LSTM, and CNN

Bioinformatics Project - B.Sc. in Computer Science and Engineering (CSE)

Created by: - Md. Tarek Hasan, Mohammed Jawwadul Islam, Md Fahad Al Rafi, Arifa Akter, Sumayra Islam

Date of Completion: - Fall 2021 Trimester (Nov 2021 - Jan 2022)

Linkedin of Jawwadul

Linkedin of Tarek

Linkedin of Fahad

Linkedin of Arifa

Linkedin of Sumayra

Code and Resources Used

  • Python Version: 3.7.11
  • Packages: numpy, pandas, keras, tensorflow, sklearn
  • Dataset from: Nguyen who is one the authors of the paper titled “DNA sequence classification by convolutional neural network”

Features of the Dataset

DNA sequences wrapped around histone proteins are the subject of datasets

  • For our experiment, we selected one of the datasets entitled H3K4me2.
  • H3K4me2 has 30683 DNA sequences whose 18143 samples fall under the positive class, the rest of the samples fall under the negative class, and it makes the problem binary class classification.
  • The ratio of the positive-negative class is around (59:41)%.
  • The class label represents the presence of H3K4me2 histone proteins in the sequences.
  • The base length of the sequences is 500.

Data Preprocessing

  • The datasets were gathered in.txt format. We discovered that the dataset contains id, sequence, and class label during the Exploratory Data Analysis phase of our work.
  • We dropped the id column from the dataset because it is the only trait that all of the samples share.
  • Except for two samples, H3K4me2 includes 36799 DNA sequences, the majority of which are 500 bases long. Those two sequences have lengths of 310 and 290, respectively. To begin, we employed the zero-padding strategy to tackle the problem. However, because there are only two examples of varying lengths, we dropped those two samples from the dataset later for experiments, as these samples may cause noise.
  • we have used the K-mer sequence representation technique to represent a DNA sequence, we have used the K-mer sequence representation technique
  • For sequence emdedding after applying the 3-mer representation technique, we have experimented using different embedding techniques. The first three embedding methods are named SequenceEmbedding1D, SequenceEmbedding2D, SequenceEmbedding2D_V2, Word2Vec and BERT.
    • SequenceEmbedding1D is the one-dimensional representation of a single DNA sequence which is basically the one-hot encoding.
    • SequenceEmbedding2D is the two-dimensional representation of a single DNA sequence where the first row is the one-hot encoding of a sequence after applying 3-mer representation. The second row is the one-hot encoding of a left-rotated sequence after applying 3-mer representation.
    • the third row of SequenceEmbedding2D_V2 is the one-hot encoding of a right-rotated sequence after applying 3-mer representation.
    • Word2Vec and BERT are the word embedding techniques for language modeling.

Deep Learning Models

After the completion of sequence embedding, we have used deep learning models for the classification task. We have used two different deep learning models for this purpose, one is Convolutional Neural Network (CNN) and the other is Bidirectional Long Short-Term Memory (Bi-LSTM).

Experimental Analysis

After the data cleaning phase, we had 36797 samples. We have used 80% of the whole dataset for training and the rest of the samples for testing. The dataset has been split using train_test_split from sklearn.model_selection stratifying by the class label. We have utilized 10% of the training data for validation purposes. For the first five experiments we have used batch training as it was throwing an exception of resource exhaustion.

The evaluation metrics we used for our experiments are accuracy, precision, recall, f1-score, and Matthews Correlation Coefficient (MCC) score. The minimum value of accuracy, precision, recall, f1-score can be 0 and the maximum value can be 1. The minimum value of the MCC score can be -1 and the maximum value can be 1.

image

Discussion

MCC score 0 indicates the model's randomized predictions. The recall score indicates how well the classifier can find all positive samples. We can say that the model's ability to classify all positive samples has been at an all-time high over the last five experiments. The highest MCC score we received was 0.1573, indicating that the model is very near to predicting in a randomized approach. We attain a maximum accuracy of 60.27%, which is much lower than the state-of-the-art result of 71.77%. To improve the score, we need to emphasize more on the sequence embedding approach. Furthermore, we can experiment with various deep learning techniques.

You might also like...
DNA-RECON { Automatic Web Reconnaissance Tool }
DNA-RECON { Automatic Web Reconnaissance Tool }

ABOUT TOOL : DNA-RECON is an automatic web reconnaissance tool written in python. This tool made for reconnaissance and information gathering with an

A repository with exploration into using transformers to predict DNA ↔ transcription factor binding

Transcription Factor binding predictions with Attention and Transformers A repository with exploration into using transformers to predict DNA ↔ transc

E2EDNA2 - An automated pipeline for simulation of DNA aptamers complexed with small molecules and short peptides

E2EDNA2 - An automated pipeline for simulation of DNA aptamers complexed with small molecules and short peptides

A script written in Python that returns a consensus string and profile matrix of a given DNA string(s) in FASTA format.

A script written in Python that returns a consensus string and profile matrix of a given DNA string(s) in FASTA format.

Pytorch Implementation of Adversarial Deep Network Embedding for Cross-Network Node Classification

Pytorch Implementation of Adversarial Deep Network Embedding for Cross-Network Node Classification (ACDNE) This is a pytorch implementation of the Adv

RefineGNN - Iterative refinement graph neural network for antibody sequence-structure co-design (RefineGNN)

Iterative refinement graph neural network for antibody sequence-structure co-des

The official implementation of the IEEE S&P`22 paper "SoK: How Robust is Deep Neural Network Image Classification Watermarking".

Watermark-Robustness-Toolbox - Official PyTorch Implementation This repository contains the official PyTorch implementation of the following paper to

Selene is a Python library and command line interface for training deep neural networks from biological sequence data such as genomes.
Selene is a Python library and command line interface for training deep neural networks from biological sequence data such as genomes.

Selene is a Python library and command line interface for training deep neural networks from biological sequence data such as genomes.

Hl classification bc - A Network-Based High-Level Data Classification Algorithm Using Betweenness Centrality
Hl classification bc - A Network-Based High-Level Data Classification Algorithm Using Betweenness Centrality

A Network-Based High-Level Data Classification Algorithm Using Betweenness Centr

Owner
Mohammed Jawwadul Islam Fida
CSE student. Founding Vice President of Students' International Affairs Society at CIAC, UIU
Mohammed Jawwadul Islam Fida
Using deep learning to predict gene structures of the coding genes in DNA sequences of Arabidopsis thaliana

DeepGeneAnnotator: A tool to annotate the gene in the genome The master thesis of the "Using deep learning to predict gene structures of the coding ge

Ching-Tien Wang 3 Sep 9, 2022
Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers

Segmentation Transformer Implementation of Segmentation Transformer in PyTorch, a new model to achieve SOTA in semantic segmentation while using trans

Abhay Gupta 161 Dec 8, 2022
Implementation of SETR model, Original paper: Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers.

SETR - Pytorch Since the original paper (Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers.) has no official

zhaohu xing 112 Dec 16, 2022
Understanding and Improving Encoder Layer Fusion in Sequence-to-Sequence Learning (ICLR 2021)

Understanding and Improving Encoder Layer Fusion in Sequence-to-Sequence Learning (ICLR 2021) Citation Please cite as: @inproceedings{liu2020understan

Sunbow Liu 22 Nov 25, 2022
[CVPR 2021] Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers

[CVPR 2021] Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers

Fudan Zhang Vision Group 897 Jan 5, 2023
Sequence to Sequence Models with PyTorch

Sequence to Sequence models with PyTorch This repository contains implementations of Sequence to Sequence (Seq2Seq) models in PyTorch At present it ha

Sandeep Subramanian 708 Dec 19, 2022
Sequence-to-Sequence learning using PyTorch

Seq2Seq in PyTorch This is a complete suite for training sequence-to-sequence models in PyTorch. It consists of several models and code to both train

Elad Hoffer 514 Nov 17, 2022
Pervasive Attention: 2D Convolutional Networks for Sequence-to-Sequence Prediction

This is a fork of Fairseq(-py) with implementations of the following models: Pervasive Attention - 2D Convolutional Neural Networks for Sequence-to-Se

Maha 490 Dec 15, 2022
Sequence lineage information extracted from RKI sequence data repo

Pango lineage information for German SARS-CoV-2 sequences This repository contains a join of the metadata and pango lineage tables of all German SARS-

Cornelius Roemer 24 Oct 26, 2022
Official repository of OFA. Paper: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework

Paper | Blog OFA is a unified multimodal pretrained model that unifies modalities (i.e., cross-modality, vision, language) and tasks (e.g., image gene

OFA Sys 1.4k Jan 8, 2023