Automatic classification and segmentation of single-molecule fluorescence time traces with deep learning

Jieming Li,Leyou Zhang,Nils G Walter,Alexander Johnson-Buck

doi:10.1038/s41467-020-19673-1

Jieming Li, Leyou Zhang + Show 2 more

Open Access

PDF Available

https://doi.org/10.1038/s41467-020-19673-1

Copy DOI

Export

Save

Cite

Abstract
Highlights/Summary
Full-Text PDF
Similar Papers

Abstract

Listen

Traces from single-molecule fluorescence microscopy (SMFM) experiments exhibit photophysical artifacts that typically necessitate human expert screening, which is time-consuming and introduces potential for user-dependent expectation bias. Here, we use deep learning to develop a rapid, automatic SMFM trace selector, termed AutoSiM, that improves the sensitivity and specificity of an assay for a DNA point mutation based on single-molecule recognition through equilibrium Poisson sampling (SiMREPS). The improved performance of AutoSiM is based on accepting both more true positives and fewer false positives than the conventional approach of hidden Markov modeling (HMM) followed by hard thresholding. As a second application, the selector is used for automated screening of single-molecule Förster resonance energy transfer (smFRET) data to identify high-quality traces for further analysis, and achieves ~90% concordance with manual selection while requiring less processing time. Finally, we show that AutoSiM can be adapted readily to novel datasets, requiring only modest Transfer Learning.

Highlights

Traces from single-molecule fluorescence microscopy (SMFM) experiments exhibit photophysical artifacts that typically necessitate human expert screening, which is time-consuming and introduces potential for user-dependent expectation bias
As a challenging data analysis case that provides a useful experimental ground truth against which to judge the performance of the algorithm, we tested the potential of deep learning to improve the analysis of single-molecule recognition through equilibrium Poisson sampling (SiMREPS) data for detection of the EGFR mutation T790M18 (Fig. 2)
Since the most timeconsuming tasks in single-molecule Förster resonance energy transfer (smFRET) for most laboratories are the curation, classification, and segmentation of smFRET time traces (Fig. 1), we developed a deep learning-based component of AutoSiM that automates these steps for two-channel smFRET traces

Summary

Introduction

Traces from single-molecule fluorescence microscopy (SMFM) experiments exhibit photophysical artifacts that typically necessitate human expert screening, which is time-consuming and introduces potential for user-dependent expectation bias. We use deep learning to develop a rapid, automatic SMFM trace selector, termed AutoSiM, that improves the sensitivity and specificity of an assay for a DNA point mutation based on single-molecule recognition through equilibrium Poisson sampling (SiMREPS). SmFRET is a widely used technique to measure small-scale distance changes (typically in the range of ~2 to 8 nm) by detecting changes in the efficiency of FRET over time in each molecule or complex. This approach permits the observation of equilibrium biomolecular dynamics that would be inaccessible to ensemble techniques[5]. An intensity-versus-time trace is generated for each candidate molecule, reducing the size of a typical dataset to ~10–100 MB/movie

Methods

Results

Conclusion