Pediatric_Brain_Cancer_Classification

Introduction

This repository utilizes generative AI capabilities by leveraging OpenAI's GPT-3.5 model, with bioinformatics analyses focused on pediatric brain cancer research. The analyses are based on the Pediatric Brain Cancer study published in Cell in 2020. Research Article: Pediatric Brain Cancer (CPTAC/CHOP, Cell 2020)

Data Description

The data used in this analysis is sourced from cBioportal and is derived from the Pediatric Brain Cancer study. Here are the details:

• Source: cBioportal
• Sample size: 218
• Classes: medulloblastoma, ganglioglioma, low grade glioma, ependymoma, high grade glioma, craniopharyngioma, and atypical teratoid rhabdoid tumor.
• Purpose: Samples are selected from the MBL and GNG classes, focusing on 2000 genes for demonstration purposes.

Notes:

• Medulloblastoma (MBL): The most common malignant brain tumor in pediatric patients, comprising 15% of all pediatric CNS tumors. It arises from the cerebellum.
• Gangliogliomas (GBL): Benign tumors of the CNS that can occur anywhere in the CNS, most commonly found in the temporal lobe.

Objectives

The primary objective of this analysis is to identify genes with the potential to discriminate between MBL and GNG classes.

Analyses

1. Differential Expression Analysis:

   • Utilize generative AI functions to identify differentially expressed genes using t-tests between the MBL and GNG classes, assuming normalized data.
   • Implement functions to conduct t-tests and calculate log fold changes for each gene, returning results as dataframes.

2. Principal Component Analysis (PCA):

   • Conduct PCA with differentially expressed genes.
   • Visualize PC1 and PC2, coloring samples based on the categorical variable using Seaborn.

3. Enrichment Analysis:

   • Perform enrichment analysis to identify pathways enriched by differentially expressed genes.
   • Implement enrichment analysis using the gseapy package.

4. Machine Learning Classifier:

   • Build a machine learning-based classifier using genes obtained from differential expression.
   • Utilize logistic regression for binary classification of MBL and GNG.
   • Split data into 70% training and 30% test sets.
   • Evaluate classifier performance metrics including accuracy, AUROC, precision, recall, and plot the ROC curve.