Drug–Target Interaction Predictor
Overview:
This project builds a basic machine learning model to predict drug–target interactions using IC50 bioactivity data. The task is treated as a binary classification problem where interactions are labeled as active or inactive based on a defined threshold.

Dataset:
The dataset (interactions.tsv) contains:
Ligand SMILES – chemical representation of the drug
Target Name – protein target
IC50 (nM) – bioactivity value
For efficient processing, the first 20,000 rows are used.

Workflow:
1. Data Loading
Loaded TSV dataset using pandas
Limited rows for performance

2. Column Selection
Selected only required columns
Ligand SMILES
Target Name
IC50 (nM)

3. Data Cleaning
Removed missing values
Removed duplicate entries
Converted IC50 to numeric format

4. Label Creation
Active (1) → IC50 ≤ 1000 nM
Inactive (0) → IC50 > 1000 nM

5. Sampling
If dataset size exceeds 15,000 rows, random sampling is applied

6. Feature Encoding
Encoded drug SMILES using LabelEncoder
Encoded target names using LabelEncoder

7. Train-Test Split
80% training data
20% testing data
Stratified sampling applied

8. Model Training
Used Random Forest classifier
Trained on encoded drug and target features

9. Model Evaluation
Calculated Accuracy score
Generated Classification Report
Precision
Recall
F1-score

Output Files:
prediction_results.csv
Contains predicted and actual interaction labels
Used to compare model predictions with true values
model_report.txt
Contains accuracy and detailed classification metrics
Console output
Displays model performance after training

Technologies Used:
Python
pandas
NumPy
scikit-learn