Machine Learning Based Diagnosis of Inflammatory Bowel Disease Using Liquid Biopsies

Table of Contents

1. Data Preprocessing
2. Data Integration
3. Differential Gene Expression Analysis
4. Functional Annotation
5. Biomarker Panel Discovey
6. Real-life Cohort Panel Validation

Step 1: Data Preprocessing

In this step, by using the preprocessing script, the microarray expression input data was preprocessed.

Step 2: Data Integration

Using the batch_integration script, the datasets were integrated to create a uniform metadata for expression analysis and batch effects removal process was followed. Additionally, genes that did not have significant signals in microarray were removed to ensure the capturability of biomarkers in experimental procedures.

Step 3: Differential Gene Expression Analysis

Bulk transcriptomic profiles of cases and controls were compared using the differential_expression_analysis script.

Step 4: Functional Annotation

By using the functional_annotation script, functional enrichment analysis and network analysis were conducted.

Step 5: Biomarker Panel Discovery

Using the biomarker_panel_discovery script, IBD-specific diagnsotic were identified and 20/80 split classification was performed using support vector machine algorithm.

Step 6: Real-life Cohort Panel Validation

The developed diagnostic biomarker panel was evaluated using the realLife_cohort_evaluation script to classify patients in the real-life cohort with qRT-PCR expression data.

For further explanation on the methodology please refer to doi: