Automatic classification of single-molecule force spectroscopy traces from heterogeneous samples.

Area of Science:

• Biophysics
• Computational Biology
• Biochemistry

Background:

• Single-molecule force spectroscopy (SMFS) generates complex datasets of protein unfolding traces.
• Analyzing heterogeneous samples with mixed protein compositions presents significant challenges for data interpretation.
• A need exists for automated methods to distinguish and classify protein unfolding patterns within complex SMFS data.

Purpose of the Study:

• To develop and present an automated data analysis pipeline for recognizing recurrent patterns (clusters) in SMFS protein unfolding traces.
• To demonstrate the pipeline's capability in handling datasets from heterogeneous biological samples.
• To extract meaningful information from noisy SMFS data without prior knowledge of sample composition.

Main Methods:

• Introduction of a novel data analysis pipeline designed for SMFS trace analysis.
• Application of the pipeline to large datasets, including tandem GB1 protein and native rod membrane samples.
• Utilizing pattern recognition and clustering algorithms to identify distinct unfolding signatures.

Main Results:

• Successful identification of multiple distinct unfolding clusters within complex SMFS datasets.
• Demonstrated effectiveness on datasets with approximately 50,000 and 400,000 traces.
• The method performs robustly despite a low signal-to-noise ratio in the experimental data.

Conclusions:

• The developed automated pattern classification pipeline effectively extracts relevant information from heterogeneous SMFS samples.
• This approach allows for the analysis of complex biological samples without requiring prior knowledge of their composition.
• The pipeline offers a valuable tool for advancing the analysis of single-molecule biophysics data.