DNA Fragment Length Analysis Using Machine Learning Assisted Vibrational Spectroscopy.
4 affiliations
- 1Optoelectronics Research Centre, Building 53, University of Southampton, University Road, Southampton, Hampshire SO17 1BJ, United Kingdom.
- 2Perioperative and Critical Care Theme, NIHR Southampton Biomedical Research Centre, Southampton Centre for Biomedical Research, University Hospital Southampton NHS Foundation Trust, Tremona Road, Southampton, Hampshire SO16 6YD, United Kingdom.
- 3The Breast Cancer Now Toby Robins Research Centre, The Institute of Cancer Research, Mary-Jean Mitchell Green Building, Chester Beatty Laboratories, 237 Fulham Road, London, SW3 6JB, United Kingdom.
- 4The Royal Marsden NHS Foundation Trust, Fulham Road, London, SW3 6JJ, United Kingdom.
Abstract
DNA length analysis is essential for genomic workflows including next-generation sequencing and fragmentomics-based diagnostics. Conventional approaches typically require large, expensive instrumentation and sample-destructive protocols with long processing times. Here we present a rapid, label-free approach integrating vibrational spectroscopy with deep learning to quantify DNA fragment length distributions. We demonstrate that ATR-FTIR and Raman spectroscopy capture length-dependent spectral features arising from phosphate backbone, nucleobase, and structural vibrations. Machine learning models trained on spectra acquired from purified monodisperse DNA (50-300 bp) predicted DNA length with high accuracy (R ² = 0.92-0.94), with multimodal fusion improving performance to R ² = 0.96. A convolutional neural network trained on DNA mixtures comprising molecules of different lengths also successfully recovered their fragment length profile. Transfer learning enabled adaptation to biological samples, achieving low prediction error (RMSE = 0.18-7.2%, Δμ = 12 bp). Importantly, the method requires only 4 μL of sample and 15 min of passive drying, with no consumables beyond cleaning materials, and allows full sample recovery. These results demonstrate the initial feasibility of vibrational spectroscopy as a rapid, label-free, and nondestructive approach for estimating DNA fragment length distributions.
Related
- Vibrational Spectroscopy in the Study of Reaction Mechanism
- Machine learning–enhanced label-free Raman spectroscopy for accurate serum-based detection of early-stage breast cancer via tumor-associated antigens
- Machine-learning-enhanced Optical Photothermal infrared (O-PTIR) spectroscopy for rapid simultaneous identification of multiple metabolites to monitor biotransformations: proof of concept on Brewer's spent grain (BSG)
- Vibrational spectroscopy and artificial intelligence for dynamic phenotypic characterization of antimicrobial resistance
- Simultaneous CEA and CDH-17 Detection from Cancer Extracellular Vesicles Via Machine Learning Enabled Bio Silent Azide-Infrared Spectroscopy
BibTeX
@article{fatayer2026fragment,
title = {{DNA Fragment Length Analysis Using Machine Learning Assisted Vibrational Spectroscopy.}},
author = {Rashad Fatayer and Waseem Ahmed and Irene Szeto and Stephen-John Sammut and Ganapathy Senthil Murugan},
journal = {ACS measurement science au},
year = {2026},
month = aug,
doi = {10.1021/acsmeasuresciau.6c00099},
url = {https://europepmc.org/article/MED/42921424},
note = {CC BY},
}
RIS
TY - JOUR
TI - DNA Fragment Length Analysis Using Machine Learning Assisted Vibrational Spectroscopy.
AU - Rashad Fatayer
AU - Waseem Ahmed
AU - Irene Szeto
AU - Stephen-John Sammut
AU - Ganapathy Senthil Murugan
PY - 2026
DA - 2026/08/10
JO - ACS measurement science au
DO - 10.1021/acsmeasuresciau.6c00099
UR - https://europepmc.org/article/MED/42921424
AB - DNA length analysis is essential for genomic workflows including next-generation sequencing and fragmentomics-based diagnostics. Conventional approaches typically require large, expensive instrumentation and sample-destructive protocols with long processing times. Here we present a rapid, label-free approach integrating vibrational spectroscopy with deep learning to quantify DNA fragment length distributions. We demonstrate that ATR-FTIR and Raman spectroscopy capture length-dependent spectral features arising from phosphate backbone, nucleobase, and structural vibrations. Machine learning models trained on spectra acquired from purified monodisperse DNA (50-300 bp) predicted DNA length with high accuracy (R ² = 0.92-0.94), with multimodal fusion improving performance to R ² = 0.96. A convolutional neural network trained on DNA mixtures comprising molecules of different lengths also successfully recovered their fragment length profile. Transfer learning enabled adaptation to biological samples, achieving low prediction error (RMSE = 0.18-7.2%, Δμ = 12 bp). Importantly, the method requires only 4 μL of sample and 15 min of passive drying, with no consumables beyond cleaning materials, and allows full sample recovery. These results demonstrate the initial feasibility of vibrational spectroscopy as a rapid, label-free, and nondestructive approach for estimating DNA fragment length distributions.
SN - 2694-250X
ER -