Classification with Missing Data - A NIFty Pipeline for Single-Cell Proteomics.

Area of Science:

• Proteomics
• Computational Biology
• Machine Learning

Background:

• Single-cell proteomics (SCP) is crucial for cell-type characterization, trajectory inference, and microenvironment mapping.
• Accurate cell-type annotation in SCP often relies on machine learning, but current methods face challenges with missing data, circular analysis, and batch effects.
• These limitations hinder the reliability and comparability of SCP data analysis.

Purpose of the Study:

• To develop a robust machine learning classification pipeline for single-cell proteomics.
• To address statistical and computational disadvantages of existing SCP annotation methods.
• To improve the accuracy and applicability of cell-type classification in single-cell proteomics experiments.

Main Methods:

• A novel top-scoring pairs based feature selection method, NIFty, was developed.
• NIFty was implemented in a full classification pipeline for single-cell proteomics data.
• The method was evaluated on datasets with missing values, batch effects, and for multiclass classification tasks.

Main Results:

• NIFty successfully classified cell types without requiring pre-imputed data.
• The method demonstrated robustness against significant batch effects without explicit correction.
• NIFty achieved comparable or superior classification accuracy compared to existing methods across varied datasets.

Conclusions:

• NIFty offers a significant advancement in single-cell proteomics data analysis by overcoming key computational and statistical hurdles.
• The method enhances the reliability of cell-type classification, enabling more accurate biological hypothesis evaluation.
• NIFty provides a powerful tool for researchers working with complex single-cell proteomics datasets.