Enhanced cancer classification and critical feature visualization using Raman spectroscopy and convolutional neural...

Area of Science:

• Biotechnology
• Spectroscopy
• Bioinformatics

Background:

• Cell line misidentification and cross-contamination compromise research integrity and resource allocation.
• Traditional cell line identification methods are time-consuming and labor-intensive.
• There is a critical need for rapid, automated cell line identification techniques.

Purpose of the Study:

• To develop and validate a novel method for rapid and accurate cell line identification.
• To assess the efficacy of Raman spectroscopy combined with a Sparrow Search Algorithm-Convolutional Neural Network (SSA-CNN) for distinguishing between normal and cancer cell lines.
• To explore the potential of SSA-CNN for biomarker discovery through visualization of spectral features.

Main Methods:

• Utilized Raman spectroscopy for label-free, non-invasive molecular analysis of cell lines.
• Developed a Sparrow Search Algorithm-Convolutional Neural Network (SSA-CNN) model for cell line classification.
• Analyzed both full spectra and fingerprint regions for enhanced identification accuracy.
• Employed Gradient-weighted Class Activation Mapping (Grad-CAM) to visualize key Raman spectral features.

Main Results:

• The SSA-CNN model achieved high accuracy (around 95%) and low standard error (≤3%) in distinguishing between six cell lines (one normal, five cancer).
• The model performed effectively using both full spectra and fingerprint regions.
• Grad-CAM visualization identified common biomolecules and specific feature peaks aligning with known biomarkers.
• The method demonstrated successful classification and potential for novel biomarker identification.

Conclusions:

• The proposed SSA-CNN strategy offers a rapid, accurate, and automated solution for cell line identification.
• Raman spectroscopy coupled with SSA-CNN enhances research efficiency by overcoming limitations of traditional methods.
• This approach serves as a valuable tool for both cell line authentication and the discovery of new cancer biomarkers.