Machine-Learning-in-Bioinformatics-using-WEKA

Used WEKA machine learning tool to analyze the Breast cancer data and also study Protein characteristics using SMO, naïve byes and IBk models also, data bias was removed using bagging and stacking. R program was used to extract the character-specific protein sequences using the R package PROTr and peptide.

Protein Feature Extraction & Breast Cancer ML Analysis

This repository contains two main components related to bioinformatics and machine learning:

1. Protein Feature Extraction using R
2. Breast Cancer Classification using Machine Learning (Weka)

---

📁 Files Included

• protein_feature_extraction.R: R script to extract multiple features from protein sequences using protr, Peptides, and foreign packages. Outputs are saved as .csv and .arff files for use in machine learning.
• JYOTHI_SWAROOP_ML_ASSIGNMENT.pdf: A detailed machine learning assignment using the Breast Cancer Wisconsin dataset analyzed through Weka.

---

🔬 1. Protein Feature Extraction (R Script)

The R script performs the following:

• Reads protein sequences from FASTA files.
• Filters valid sequences using protcheck.
• Extracts various protein descriptors using:
  • extractDC, extractTC, extractCTDC, extractCTDD, extractCTDT, extractAPAAC, extractAAC
• Saves feature matrices for each descriptor in .csv format.
• Combines positive/negative classes and exports them as .arff files for ML tools like Weka.

Packages used:

• protr
• Peptides
• foreign

---

🧠 2. Machine Learning Assignment (Weka)

The assignment explores classification models applied to the Breast Cancer Wisconsin dataset. The goal is to predict tumor recurrence based on morphological features.

🧪 Models Evaluated:

• IBk (k-Nearest Neighbors) – Best performer
• Naïve Bayes
• Support Vector Machine (SMO)

🔍 Key Findings:

Model	Accuracy	Recall	F1-Score	ROC AUC	PRC AUC
IBk (k=1)	72.37%	0.724	0.697	0.628	0.686
Naïve Bayes	71.67%	0.717	0.708	0.701	0.741
SMO (SVM)	69.58%	0.696	0.600	0.509	0.586

• Best overall: IBk (k-NN), especially for recall—critical in medical diagnosis.
• Naïve Bayes handled noise and missing values well.
• SVM underperformed due to data characteristics and model assumptions.

---

🧬 Biological Relevance

• Machine learning helps predict recurrence risk, aiding early diagnosis and personalized treatment.
• Protein features extracted can be applied to similar predictive modeling tasks.
• Bioinformatics + ML provides a powerful toolset for computational biology.

---

✅ How to Use

For R Script:

1. Ensure required R packages are installed: protr, Peptides, foreign.
2. Update file paths in the script if needed.
3. Run the script to generate .csv and .arff files for your sequences.

For Weka Assignment:

1. Open Weka GUI.
2. Load the .arff files.
3. Apply classifiers under the "Classify" tab.
4. Evaluate results using 10-fold cross-validation.

---

📌 Author

Jyothi Swaroop C
Machine Learning in Bioinformatics – Assignment & R-based Feature Extraction

---

📬 Contact

Feel free to reach out if you have any questions or want to collaborate!