Raman and SERS analytical platform with machine learning for classifying wild type p53 and hotspot mutants R175H and...

Area of Science:

• Biophysics
• Analytical Chemistry
• Biomedical Engineering

Background:

• Detecting subtle protein conformational changes is crucial for disease diagnostics and biomolecular characterization.
• The tumor suppressor p53 protein and its mutants are important in cancer research.
• Existing methods may lack the sensitivity to distinguish closely related protein variants.

Purpose of the Study:

• To develop and evaluate an analytical platform for classifying wild-type p53 protein and its hotspot mutants.
• To assess the efficacy of combining Raman spectroscopy, SERS, and machine learning for this task.
• To demonstrate a versatile framework for studying protein conformational changes.

Main Methods:

• Utilized Raman spectroscopy and surface-enhanced Raman spectroscopy (SERS) with various nanostructured substrates (gold nanospheres, gold nanorods, silver nanoparticles on Al).
• Acquired label-free spectra of recombinant wild-type p53 and its R175H and R273H mutants.
• Employed Principal Component Analysis (PCA) for spectral fingerprinting and supervised machine learning (Linear-SVM) for classification with rigorous cross-validation.

Main Results:

• Distinct spectral fingerprints were identified for p53 variants, particularly in amide III and CH stretching regions, indicating subtle conformational differences.
• A linear Support Vector Machine (Linear-SVM) model achieved a high classification accuracy of 92.9 ± 6.9% on gold nanosphere-coated aluminum substrates (AuNS@Al).
• The integrated platform demonstrated robustness in structurally classifying wild-type and mutant p53 proteins.

Conclusions:

• The integration of optimized SERS substrates, Raman spectroscopy, and machine learning provides a powerful analytical platform for protein variant classification.
• This approach offers a versatile framework for investigating conformational changes in biomedically relevant biomolecules.
• The study highlights potential applications in cancer biomarker detection and the study of protein misfolding diseases.