Protection-oriented screening
Explore one frozen national sequence and its localized applications
COMPASS uses censored protective-apical evidence to estimate measured-tail probabilities and assemble compact species panels with complementary chemical coverage. The explorer presents the frozen Top-20, a complete 2,144-species continuation with cumulative expected capture for every prefix, evidence volume, and state-level localization.
The national sequence was constructed at the lower-5% target and is evaluated at broader targets without reoptimizing panel membership.
Frozen downloads
Download full-precision CSV files generated in the same export as the page data.
National Sequence Explorer
Complete national ordering and cumulative expected capture
Ranks 1–20 reproduce the frozen R1 strict focal-species leave-one-out analysis. Ranks 21–2,144 are a deterministic post hoc continuation under the same R1 objective; they support exploration but have not been validated as equivalently resolved biological rankings.
How evidence volume is defined: the tier counts distinct valid protective analysis contexts for a species: Very limited = 1–4, Limited = 5–19, Moderate = 20–49, and Strong = 50 or more. The adjacent “Chemicals” value is the number of unique chemicals represented. These fields describe evidence volume; sensitivity is evaluated separately by the model.
— ranked species match the current filters.
| Rank | Species | Taxonomy | Protective evidence volume | Contexts | Chemicals | Greedy gain | Cumulative expected capture | Scope |
|---|
—
Full-sequence cumulative capture
All 2,144 prefixes of one fixed sequence; the panel-size axis is logarithmic so early and late prefixes remain visible.
Species Catalog
Protective evidence volume across the formal candidate universe
Counts summarize observed protective evidence volume and occurrence relevance. All 2,144 formal-binomial candidates have a national order; species with limited evidence remain visible and are marked for confirmatory testing.
Evidence-volume tiers: Very limited = 1–4 protective contexts; Limited = 5–19; Moderate = 20–49; Strong = ≥50. “Protective chemicals” counts unique chemicals and may be smaller than the context count.
— species match the current filters.
| Species | National rank | Taxonomy | Protective contexts | Protective chemicals | Warning contexts | Eligible states | Protective evidence volume | Note |
|---|
—
Regional Localization Explorer
State-priority weighting at the lower-5% target
Regional panels were optimized at the lower-5% target.
Select a state on the map
Click a state or use the menu below. Hover or focus a state to view its Top-5 basis and species.
Selected state
Boundaries: U.S. Census Bureau 2024 1:500,000 cartographic state boundaries. Alaska and Hawaii are shown as insets.
Localized Top-5
Fixed national Top-5
| Rank | Species | Candidate status | Species relevance | Mean tail probability | Mean reliability | Direct-support chemicals |
|---|
Methods & Data
What the explorer shows
Measured-tail probability
The calibrated probability that a species threshold lies within a measured lower-tail target for the stated chemical and protective effect context. Exact and bounded observations enter a censored likelihood.
Expected capture
The chemical-priority-weighted probability that at least one selected panel member occupies the working measured lower tail. Values are stored as probabilities and displayed as percentages.
Complementarity
Species are combined because their supported chemical domains can complement one another. A species' incremental gain describes its addition to the fixed sequence at the lower-5% target.
National sequence
One U.S.-wide sequence is selected at the lower-5% target from R1 strict focal-species leave-one-out probabilities. Ranks 1–20 reproduce the frozen R1 analysis; all 2,144 fixed-sequence prefixes are evaluated at the lower-5%, lower-10%, and lower-20% targets with the R1 target-specific working-dependence model.
Protective evidence volume
Evidence-volume bands count distinct valid protective analysis contexts: 1–4, 5–19, 20–49, or at least 50. The band describes how much direct evidence supports a species estimate; sensitivity is evaluated separately by the model.
Regional localization
State-priority chemical weights are combined with soft species relevance and method-support weights. Occurrence/NAS evidence informs relevance and follow-up needs.
Top5-apical
The minimum identified protective-apical threshold among available fixed Top-5 members is compared with measured full-data reference HC5 contexts to evaluate rank information.
Chemical follow-up
The fixed-panel measured-tail probability score identifies chemicals for additional evidence acquisition. Follow-up prioritization is interpreted within its evidence and chemical context.
Warning endpoints
Heartbeat, behavior, feeding, avoidance, and physiological evidence provide response-window and rank signals that guide apical confirmation.
Data boundary
Primary inputs include large public datasets such as EPA ECOTOX and other provider-hosted resources. Their size and provider distribution arrangements make them unsuitable for mirroring in the GitHub code repository, so users can obtain them from the official sources listed in the code package and prepare the documented local inputs. This explorer contains the compact frozen derived tables needed for its precomputed modules.
Version & Citation
Frozen release metadata
Manuscript
—
Authors:
Release identifiers
GitHub release: COMPASS v1.0.1
Zenodo DOI: No DOI has been assigned to this release.
License: Publicly viewable source; all rights reserved.