End-to-end workflows for liquid biopsy biotyping analysis using combined MALDI MS and machine learning approach.

Area of Science:

• Biochemistry
• Computational Biology
• Oncology

Background:

• Liquid biopsy analysis using MALDI MS shows promise for non-invasive disease detection.
• Machine learning integration can enhance the predictive power of MALDI MS data.
• Standardized workflows are needed for clinical translation of these techniques.

Purpose of the Study:

• To present an open-source R-based workflow for end-to-end MALDI MS liquid biopsy analysis.
• To enable customizable and transparent data preprocessing and predictive model evaluation.
• To validate the workflow on clinical samples for hemato-oncological diseases.

Main Methods:

• Development of an R-based computational pipeline for MALDI MS data.
• Implementation of data preprocessing steps, including normalization and feature extraction.
• Application of machine learning models for disease classification and monitoring.
• Validation using plasma samples from hemato-oncological patients.

Main Results:

• The R workflow provides a comprehensive and customizable pipeline for MALDI MS liquid biopsy analysis.
• The workflow demonstrated successful data preprocessing and predictive model evaluation.
• Validation on clinical samples confirmed the utility for hemato-oncological patient monitoring.
• Enhanced data reproducibility was achieved through the standardized workflow.

Conclusions:

• The presented open-source workflow facilitates the integration of MALDI MS and ML for non-invasive disease screening.
• This approach enhances reproducibility and streamlines the analysis of liquid biopsies.
• The workflow serves as a foundation for adopting MALDI MS in routine clinical practice for oncology.