Denoise Stepwise Signals by Diffusion Model-Based Approach

Area of Science:

• Physical Chemistry
• Biophysics
• Signal Processing

Background:

• Stepwise signals are crucial in single-molecule detection, reflecting molecular changes but are often obscured by noise.
• Conventional denoising methods like frequency-domain filtering and Hidden Markov Models are limited for stepwise signals due to noise overlap and stationarity assumptions.

Purpose of the Study:

• To develop a novel algorithm for denoising stepwise signals in single-molecule detection.
• To improve the accuracy of signal level estimation and transition point detection in noisy stepwise signals.

Main Methods:

• Proposed the Stepwise Signal Diffusion Model (SSDM), a diffusion model-based algorithm.
• Integrated a multi-scale convolutional network with an attention mechanism for signal reconstruction.
• Generated training data using simulated stepwise signals with additive Gaussian noise.

Main Results:

• SSDM demonstrated superior performance in signal level reconstruction and transition point detection across various signal-to-noise ratios.
• The model effectively denoises stepwise signals, outperforming traditional methods.
• Validated on experimental data from single-molecule Förster Resonance Energy Transfer and nanopore DNA translocation.

Conclusions:

• SSDM offers a robust and generalizable framework for recovering stepwise signals.
• The diffusion model approach provides significant advancements for analyzing noisy stepwise data in single-molecule science and other physical systems.