Modelling of immune infiltration in prostate cancer treated with HDR-brachytherapy using Raman spectroscopy and...

Area of Science:

• Biomedical Engineering
• Computational Biology
• Oncology

Background:

• Prostate cancer features an immunosuppressive tumor microenvironment.
• Understanding immune cell infiltration is crucial for predicting disease progression and treatment response.

Purpose of the Study:

• To develop and validate a novel methodology combining Raman spectroscopy, group-and-bases-restricted non-negative matrix factorization (GBR-NMF), and machine learning.
• To accurately model immune cell densities (CD68, CD3, CD8) within prostate cancer needle-core biopsies.
• To assess changes in immune cell densities before and after high-dose-rate brachytherapy (HDR-BT).

Main Methods:

• Raman spectral acquisition and immunohistochemistry staining for CD68, CD3, and CD8 cells were performed on patient biopsies.
• Group-and-bases-restricted non-negative matrix factorization (GBR-NMF) was used to extract spectral features.
• Supervised machine learning regression models (gradient-boosted trees, elastic net) were trained using GBR-NMF scores to predict immune cell densities.

Main Results:

• The elastic net model accurately predicted the CD68/CD8 cell ratio (R²: 0.82, RMSE: 0.25).
• The gradient-boosted trees model predicted CD68 cell density with good accuracy (R²: 0.65, RMSE: 163).
• Model accuracy, defined as predictions within one standard deviation, was 11/16 for CD68/CD3 and 12/16 for CD68/CD8 models.

Conclusions:

• This study presents a novel, data-driven approach to model immune cell densities in prostate cancer.
• The methodology demonstrates the potential of combining spectroscopy and machine learning for non-invasive assessment of the tumor microenvironment.
• Accurate modeling of immune cells can serve as prognostic indicators for disease progression, particularly in the context of HDR-BT treatment.