Deep learning-enhanced vibrational spectroscopy for quantitative DNA fragment profiling

Fatayer, Rashad, Sammut, Stephen-John and Senthil Murugan, Ganapathy (2025) Deep learning-enhanced vibrational spectroscopy for quantitative DNA fragment profiling. In Optics in Health Care and Biomedical Optics XV (2025). vol. 13721, SPIE. 6 pp . (doi:10.1117/12.3076499).

Record type: Conference or Workshop Item (Paper)

Abstract

We demonstrate a deep learning approach for quantitative DNA fragment length analysis using vibrational spectroscopy. Controlled DNA mixtures spanning 50–300 bp were produced to establish distributions, enabling interpretable regression modelling. A convolutional neural network (CNN) with an attention module was trained to predict fragment proportions from spectral data. The model reliably reconstructed distributional differences, providing clear profiles that facilitate interpretation of biologically meaningful fragmentation patterns. To our knowledge, this is the first application of vibrational spectroscopy with deep learning for resolving DNA fragment length distributions, offering a rapid, label-free, and non-destructive complement to existing molecular assays.

Text

1372113 - Accepted Manuscript

Available under License Other.

Download (990kB)