{
  "abstract": "When a chemical signal recedes into noise, deciding whether a weak chemical feature is still worth observing - and where observation should stop - is usually delegated to signal-to-noise conventions, software parameters, or an analyst’s personal judgment. We propose Observation Deviation (τ), a dimensionless measure of chemical-feature observability, that turns this tacit judgment into a quantifiable analytical variable. Within a stated observation protocol, τ quantifies how much an analytical chemical feature deviates from the local profile its observation anticipates. On simulated analytical responses and real GC–MS chromatograms, three independent observer experiments (19 observers in total) order systematically along the τ scale, including the region where observer disagreement is strongest. τ also separates human-judged clear peaks from noise in a case-matched instance where raw signal-to-noise does not. We further develop TauNet - a learned representation of chemical-feature observability - showing that this observability can be learned directly from measured response profiles and, without re-calibration, transfers from controlled simulated responses to previously unseen real chromatographic measurements. The resulting model is independently implemented as a lightweight single-file application. τ thus does not replace analytical judgment, nor does it impose a universal decision threshold; rather, under a declared observation protocol, it renders chemical-feature observability measurable, learnable, and reproducible. This contrasts with existing chemical neural-network work: for peak detection, quantitative regression, or molecular-property prediction, those methods learn to map chemical signals to chemical conclusions - signal → peak/class/property - whereas τ fills an observability layer that none of them occupy, between candidate detection and chemical interpretation.",
  "abstract_shown": true,
  "affiliations": [
    "School of Electrical and Electronic Engineering",
    "Nanyang Technological University",
    "Shimadzu (Asia Pacific) Pte Ltd",
    "ChemoPower Technology Pte Ltd"
  ],
  "article_type": null,
  "arxiv": null,
  "author_total": 3,
  "authors": [
    {
      "affs": [
        0,
        1
      ],
      "name": "Zhi Xi ZHANG"
    },
    {
      "affs": [
        2
      ],
      "name": "Chun Kiang Chua"
    },
    {
      "affs": [
        3
      ],
      "name": "Hua Jun Zhang",
      "orcid": "0000-0002-3947-0750"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/31gsen1xfjp9q.bib",
  "cited_by": 0,
  "doi": "10.26434/chemrxiv.15008283/v1",
  "fulltext_url": null,
  "id": "31gsen1xfjp9q",
  "issns": [],
  "landing_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008283/v1",
  "language": null,
  "licence": "CC-BY",
  "licence_label": "CC BY",
  "licence_source": "journal",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [
    [
      "openalex",
      "1602|Spectroscopy and Chemometric Analyses"
    ]
  ],
  "openalex": "W7208762153",
  "pdf_url": "https://chemrxiv.org/doi/pdf/10.26434/chemrxiv.15008283/v1",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-03",
  "publisher": "American Chemical Society (ACS)",
  "ris": "https://research.draconitic.com/paper/31gsen1xfjp9q.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:17:17Z",
      "native_id": "10.26434/chemrxiv.15008283",
      "source": "chemrxiv",
      "url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008283/v1"
    }
  ],
  "subjects": [
    {
      "field": "chemistry",
      "origin": "openalex:1602|Spectroscopy and Chemometric Analyses",
      "subfield": "analytical"
    }
  ],
  "title": "Observation Deviation as a Quantitative Observability Measure of Weak Chemical Features in Analytical Measurements",
  "title_text": "Observation Deviation as a Quantitative Observability Measure of Weak Chemical Features in Analytical Measurements",
  "type": "preprint",
  "updated_at": "2026-10-30T12:17:17Z",
  "url": "https://research.draconitic.com/paper/31gsen1xfjp9q",
  "venue": "ChemRxiv",
  "versions": []
}