Deep-Learning-Assisted Stratification of Amyloid Beta Mutants Using Drying Droplet Patterns

Area of Science:

• Biochemistry
• Biophysics
• Computational Biology

Background:

• Predicting protein mutations is a significant challenge in biochemistry.
• Protein structure and mutations are linked to diseases like Alzheimer's and Parkinson's.
• Understanding these alterations is crucial for disease diagnosis and treatment.

Purpose of the Study:

• To develop a simple and accurate method for predicting protein mutations.
• To investigate the potential of analyzing drying droplet stains for structural information.
• To stratify amyloid beta (1-42) (Aβ42) variants and conformations using stain patterns.

Main Methods:

• Utilized polarized light microscopy to image drying droplet deposits of Aβ42 peptides.
• Employed deep-learning neural networks to analyze complex stain patterns.
• Examined Aβ42 peptides with single amino acid differences, representing hereditary Alzheimer's disease mutants.

Main Results:

• Stain patterns from drying peptide droplets were reproducible.
• Deep learning accurately stratified eight Aβ variants with >99% predictive accuracy.
• Distinct Aβ42 peptide conformations were identified with >99% accuracy.
• Minute structural differences in peptide primary and secondary structures were detected.

Conclusions:

• Drying droplet analysis is a simple yet effective method for inferring peptide structural information.
• This technique offers a scalable and accurate approach for stratifying protein alterations.
• The findings could aid in unraveling pathological signatures in neurodegenerative diseases.