Skip to content
DraconiticResearch
Search Sources API
203,203 studies31 sourcesUpdated 39 min ago
PreprintCC BY 4.0ECSarXiv

When Raman Spectral Fidelity and Bacterial Classification Accuracy Diverge

  • yonghe l
Open at sourcePDFDOICite

Abstract

Spectral error and classification error can respond differently to the same change in a Raman spectrum. We compared three derivative errors and three versions of spectral information divergence (SID) against mean squared error (MSE) using a fixed bacterial classification task. All derivative errors were less closely aligned with classification harm than MSE. SID improved alignment gap and ordering concordance only after Gaussian smoothing at 10 cm−1. In a separate test, Gaussian broadening increased MSE throughout the tested range, while mean accuracy first rose and then fell. Including broadening among the perturbations retained Wasserstein distance’s advantage over MSE and increased peak F1’s relative disadvantage. For this classifier, closer agreement with the original spectrum did not consistently imply higher accuracy.

Related

  1. SERS-CBP: A Comprehensive SERS Spectral Database of Clinically Important Bacterial Pathogens.
    Liang Wang, Xu-Xia Cui, Jia-Wei Tang +8
    ACS measurement science auPeer-reviewedCC BY-NC-NDFull textBiophysicsEurope PMC
  2. FRACTAL: A Ratiometric-based Fingerprinting Framework for Robust Spectral Classification and Ratio-based Spectroscopic Marker Discovery
    Lamyaa M. Almehmadi
    ChemRxivPreprintCC BYFull textBiophysicsChemRxiv
  3. Fast structural search for classification of gut bacterial mucin O-glycan degrading enzymes
    Mert Erden, Tyler Schult, Karin Yanagi +4
    PLOS Computational BiologyPeer-reviewedCC BY 4.0Full textEnzymology & metabolismMolecular biology & gene regulationPLOS
  4. Rod photoreceptors have a dual dependency on both aerobic glycolysis and OXPHOS and diverge metabolically from other retinal neurons
    Gabriele M. Wögenstein, Luca Ravotto, Vyara Todorova +5
    Science AdvancesPeer-reviewedCC BYMolecular biology & gene regulationCrossref
  5. Benchmarking long-read aligner-caller pipelines for structural variant detection across ten angiosperm species with PacBio high-fidelity sequencing
    Zhiliang Zhang, Jun Xu, Zhizhan Wang +9
    Genome BiologyPeer-reviewedCC BY-NC-ND 4.0Molecular biology & gene regulationCrossref

Cite

BibTeXRIS
BibTeX
@misc{l2026when,
  title = {{When Raman Spectral Fidelity and Bacterial Classification Accuracy Diverge}},
  author = {yonghe l},
  howpublished = {ECSarXiv},
  year = {2026},
  month = oct,
  doi = {10.1149/osf.io/97am3_v1},
  url = {https://osf.io/preprints/ecsarxiv/97am3_v1/},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - When Raman Spectral Fidelity and Bacterial Classification Accuracy Diverge
AU  - yonghe l
PY  - 2026
DA  - 2026/10/13
PB  - ECSarXiv
DO  - 10.1149/osf.io/97am3_v1
UR  - https://osf.io/preprints/ecsarxiv/97am3_v1/
AB  - Spectral error and classification error can respond differently to the same change in a Raman spectrum. We compared three derivative errors and three versions of spectral information divergence (SID) against mean squared error (MSE) using a fixed bacterial classification task. All derivative errors were less closely aligned with classification harm than MSE. SID improved alignment gap and ordering concordance only after Gaussian smoothing at 10 cm−1. In a separate test, Gaussian broadening increased MSE throughout the tested range, while mean accuracy first rose and then fell. Including broadening among the perturbations retained Wasserstein distance’s advantage over MSE and increased peak F1’s relative disadvantage. For this classifier, closer agreement with the original spectrum did not consistently imply higher accuracy.
ER  -