Automated registration and clustering for enhanced localization atomic force microscopy of flexible membrane proteins

Area of Science:

• Biophysics
• Microscopy
• Computational Biology

Background:

• Atomic Force Microscopy (AFM) provides near-native imaging of biomolecules but has limited lateral resolution.
• Localization Atomic Force Microscopy (LAFM) improves resolution but struggles with flexible proteins exhibiting multiple conformations.
• The SecYEG translocon is a membrane protein complex known to adopt various conformational states.

Purpose of the Study:

• To develop an unsupervised deep learning algorithm to enhance LAFM's applicability to flexible proteins.
• To simultaneously register and cluster AFM images based on protein conformation.
• To improve the resolution of individual protein conformations for systems like SecYEG.

Main Methods:

• An unsupervised deep learning algorithm was developed for image registration and clustering.
• Simulated AFM images from molecular dynamics simulations of the SecYEG translocon were used.
• The algorithm was tested for its ability to resolve distinct protein conformations.

Main Results:

• The deep learning algorithm successfully registered and clustered AFM images by protein conformation.
• Improved resolution was achieved for individual conformations of the SecYEG translocon.
• The enhanced LAFM method demonstrated increased effectiveness for flexible protein imaging.

Conclusions:

• This work introduces a novel deep learning approach to overcome LAFM limitations with flexible proteins.
• The algorithm enables more accurate imaging of biomolecules with multiple conformational states.
• This represents a significant step towards a generalized LAFM for complex biological macromolecules.