SeqFea-Learn - An intergrated python package for DNA, RNA and Protein sequecing data analysis

Include Feature extraction, Feature selection, Dimensionality Reduction, Models Construction for sequencing data.

Table of Contents

* Installation

* Data Preparation

* DNA Feature Extraction

* RNA Feature Extraction

* Protein Feature Extraction

* Feature Selection

* Dimensionality Reduction

* Feature Evaluation

Installation

The package is developed using Python 3(Python Version 3.0 or above) and it can be run on Linux operatinig system. We strongly recommand user to install Ancaonda Python 3.7 or above version to avoid installing other packages.

After installaing Anaconda, the following packages need to be installed:

1. xgboost
2. skrebate
3. lightgbm

The source code is freely available at: https://github.com/ashinandjay/FeatureSelection

To install our tool, first download the zip file manually from github, or use the code below in Unix:

cd your_folder_path
wget https://github.com/ashinandjay/FeatureSelection/archive/master.zip

Unzip the file:

unzip master.zip

Data Preparation

The DNA, RNA or protein sequence data (FASTA format) and their labels (txt format) are required for using our feature selection tool.

DNA Feature Extraction

The tool includes 16 feature extraction methods for DNA sequencing data.

DNA Extraction Method	DNA Extraction Number
Kmer	1
Reverse Compliment Kmer	2
Pseudo dinucleotide composition	3
Pseudo k-tuple nucleotide composition	4
Dinucleotide-based auto covariance	5
Dinucleotide-based cross covariance	6
Dinucleotide-based auto-cross covariance	7
Trinucleotide-based auto covariance	8
Trinucleotide-based cross covariance	9
Trinucleotide-based auto-cross covariance	10
Nucleic acid composition	11
Di-nucleotide composition	12
Tri-nucleotide composition	13
zcurve	14
monoMonoKGap	15
monoDiKGap	16

DNA_Feature_Extraction require two inputs: DNA Extraction number and DNA sequencing data.

Run DNA_Feature_Extraction.py:

DNA_Feature_Extraction.py [DNA Extraction number] [DNA sequencing data]

Example: Use kmer method to extract features from DNA sequencing data

DNA_Feature_Extraction.py 1 DNA_sequencing.txt

RNA Feature Extraction

The tool includes 12 feature extraction methods for RNA sequencing data.

RNA Extraction Method	RNA Extraction Number
Kmer	1
Reverse Compliment Kmer	2
Pseudo dinucleotide composition	3
Dinucleotide-based auto covariance	4
Dinucleotide-based cross covariance	5
Dinucleotide-based auto-cross covariance	6
Nucleic acid composition	7
Di-nucleotide composition	8
Tri-nucleotide composition	9
zcurve	10
monoMonoKGap	11
monoDiKGap	12

RNA_Feature_Extraction require two inputs: RNA Extraction number and RNA sequencing data.

Run RNA_Feature_Extraction.py:

RNA_Feature_Extraction.py [RNA Extraction number] [RNA sequencing data]

Example: Use kmer method to extract features from RNA sequencing data

RNA_Feature_Extraction.py 1 RNA_sequencing.txt

Protein Feature Extraction

The tool includes 32 feature extraction methods for Protein sequencing data.

Protein Extraction Method	Protein Extraction Number
Amino acid composition	1
Composition of k-spaced amino acid pairs	2
Dipeptide composition	3
Grouped dipeptide composition	4
Grouped tripeptide composition	5
Cojoint triad	6
k-spaced cojoint triad	7
Composition	8
Transition	9
Distribution	10
Encoding based on grouped weight	11
Auto covariance	12
Moran autocorrelation	13
Geary autocorrelation	14
Quasi-sequence-order	15
Pseudo-amino acid composition	16
Amphiphilic pseudo-amino acid composition	17
Amino Acid Composition PSSM	18
Dipeptide composition PSSM	19
Pseudo PSSM	20
Auto covariance PSSM	21
Cross covariance PSSM	22
Auto Cross covariance PSSM	23
Bigram-PSSM	24
AB-PSSM	25
Secondary structure composition	26
Accessible surface area composition	27
Torsional angles composition	28
Torsional angles bigram	29
Structural probabilities Bigram	30
Torsional angles auto-covariance	31
Structural probabilities auto-covariance	32

Protein_Feature_Extraction require two inputs: Protein Extraction number and Protein sequencing data.

Run Protein_Feature_Extraction.py:

Protein_Feature_Extraction.py [Protein Extraction number] [Protein sequencing data]

Example: Use Amino acid composition method to extract features from Protein sequencing data

Protein_Feature_Extraction.py 1 Protein_sequencing.txt

Feature Selection

Our Feature Selection tool contains 20 supervised selection methods.

Feature Selection Method	Feature Selection Number
Lasso	1
Elastic Net	2
L1-SVM	3
CHI2	4
Pearson Correlation	5
ExtraTree	6
XGBoost	7
SVM-RFE	8
LOG-RFE	9
Mutual Information	10
Minimum Redundancy Maximum Relevance	11
Joint Mutual Information	12
Maximum-Relevance-Maximum-Distance	13
ReliefF	14
Trace Ratio	15
Gini index	16
SPEC	17
Fisher Score	18
T Score	19
Information Gain	20

For using our Feacture Selection Tool, Four inputs are required:

1. Feauture selection number (See the table above)
2. Number of feature to select (how many number of feature you want)
3. Feature Vectors (Feature extraction output file)
4. Label Vectors (labels for sequencing)

Run Feature_Selection.py:

Feature_Selection.py [Feauture selection number] [Number of feature to select] [Feature Vectors] [Label Vectors]

Example: Using Lasso method to select 3 features

Feature_Selection.py 1 3 Feaute_Vectors.csv label.txt

Dimensionality Reduction

Our Feature Reduction tool contains 16 unsupervised dimensionality reduction methods.

Dimensionality Reduction Method	Feature Reduction Number
K-means	1
T-SNE	2
Principal Component Analysis	3
Kernel PCA	4
Locally-linear embedding	5
Singular Value Decomposition	6
Non-negative matrix factorization	7
Multi-dimensional Scaling	8
Independent Component Analysis	9
Factor Analysis	10
Agglomerate Feature	11
Gaussian random projection	12
Sparse random projection	13
Autoencoder	14
Gaussian Noise Autoencoder	15
Variational Autoencoder	16

For using our Feacture Reduction Tool, Three inputs are required:

1. Feauture Reduction number (See the table above)
2. Number of Clusters (how many number of Clusters you want)
3. Feature Vectors (Feature extraction output file)

Run Feature_Reduction.py:

Feature_Reduction.py [Feauture Reduction number] [Number of Clusters to select] [Feature Vectors]

Example: Using PCA method to select 3 clusters

Feature_Reduction.py 3 3 Feaute_Vectors.csv

Feature Evaluation

Feature selection method can be evaluated using 10 classifiers and 3 deep learning methods include SVM, KNN, RandomForest, LightGBM, XGBoost, Adaboost, Bagging, ExtraTree, gaussian Naïve Bayes, gradient boosting, DNN, CNN and RNN predictors. The classification accurcay comparison files (plot and table) will be generated in same folder of code.

Run Feature_Evaluation.py:

Feature_Evaluation.py [Feauture selection output] [Label Vectors]

Example: evaluating Lasso selection method

Feature_Evaluation.py Lasso.csv label.txt