RhoMax: Computational Prediction of Rhodopsin Absorption Maxima Using Geometric Deep Learning

Area of Science:

• Biophysics
• Molecular Biology
• Neuroscience

Background:

• Microbial rhodopsins (MRs) are vital photoactive proteins used in biophysics.
• Optogenetics uses genetically engineered proteins to control neural activity with light.
• Limited light penetration in tissues hinders optogenetics, especially with blue light.

Purpose of the Study:

• To address the scarcity of red-shifted opsins for improved optogenetics.
• To develop an accurate and rapid method for predicting rhodopsin absorption wavelengths.
• To overcome the limitations of experimental and current computational methods for rhodopsin characterization.

Main Methods:

• Introduced RhoMax, a structure-based geometric deep learning approach.
• Utilized AlphaFold2 for accurate modeling of rhodopsin structures from sequences.
• Trained and validated the model on a diverse dataset of microbial rhodopsins.

Main Results:

• RhoMax accurately predicted maximum absorption wavelengths from protein sequences alone.
• Achieved high precision with an accuracy of 0.03 eV on over half of the test set.
• Demonstrated the potential to significantly reduce the time for designing novel red-shifted rhodopsins.

Conclusions:

• Computational prediction of absorption maxima is feasible and accurate.
• RhoMax facilitates the design of new red-shifted microbial rhodopsins.
• This advancement promises to overcome key limitations in optogenetics research.