The EvoBNE Atlas Qualification Standard defines the computational protocol used to evaluate candidate proteins. It separates external novelty from internal Atlas relationships and characterizes each candidate across sequence, structural, physicochemical, and ensemble dimensions.
Transparent qualification. Proprietary discovery.
Qualification is a computational determination, not an experimental one. A qualified candidate was evaluated against the standard below and met its criteria. It is not a claim of biological function, material performance, or safety — those require experimental validation, which this pipeline does not substitute for.
A candidate described as computationally dark had no convincing sequence or structural relationship detected in the databases searched, under the conditions stated below. That is a statement about a search, not a proof of absolute novelty.
Each candidate is run through four largely independent analyses. No single output is decisive on its own.
Searches the candidate against public sequence and structure databases, then against every other Atlas candidate, to establish what is and isn't already known.
Scans for known biomaterial-relevant sequence signatures — silk, elastin, collagen, amyloid-adjacent, and cell-adhesion motifs — as sequence facts, not predictions.
Characterizes the candidate's static predicted structure: shape, secondary structure, disulfides, charge patterning, disorder, and an independent refolding check.
Simulates the candidate's dynamic conformational behavior under a coarse-grained physical model and classifies the resulting ensemble.
These are answered by the same tools but are not the same measurement, and should never be read as one number. Collapsing them is the single most common misreading of this methodology.
| Question asked | Search | Value | |
|---|---|---|---|
| Novelty | Does anything like this already exist in public sequence/structure space? | blastp vs. nr; sequence search vs. RCSB PDB and UniParc; structure search vs. PDB100, AlphaFold DB, ESM Atlas | E-value ≤ 10 (permissive, sensitivity-first); TM-score > 0.5 = related |
| Atlas homology | Is this candidate part of a family within the Atlas cohort itself? | all-vs-all blastp and TM-align across Atlas candidates only | E ≤ 1×10⁻⁵, identity ≥ 30%; TM-score > 0.5 |
Novelty search is deliberately permissive — a loose E-value casts the widest possible net so nothing similar is missed. Atlas homology classification is deliberately strict — a tight E-value ensures two candidates are only grouped as a family when the relationship is statistically unambiguous. The same E-value threshold would be the wrong choice for the other question.
A general sense of what to expect in a candidate's qualification record — not an exhaustive list.
| Pipeline | Representative properties |
|---|---|
| Homology & Novelty | TM-score, sequence identity/coverage, family assignment |
| Motif Scanner | Motif counts and positions (silk, elastin, collagen, amyloid-adjacent, cell-adhesion) |
| Structural & Physicochemical | Secondary structure %, shape (asphericity, κ²), disulfide pairs, net charge, GRAVY, isoelectric point (pI), charge patterning (FCR, NCPR, κ), disorder score |
| EDAAP / Ensemble | Contact frequencies, radius-of-gyration distribution, polymer scaling exponent (ν), conformational basins/clusters, per-residue flexibility (RMSF) |
Qualification is determined from three complementary components of the standard. No single descriptive property is treated as a substitute for the complete analysis.
The candidate must be computationally dark. This is a hard gate, not a factor weighed against others.
Structural and physicochemical properties are read together — as a pattern, not a formula — for biomaterial-application potential.
Simulation-derived evidence is only trusted once the underlying simulation is confirmed to have converged.
The methodology is applied as a fixed analytical standard rather than a candidate-specific or application-specific threshold. Tools, databases, parameters, thresholds, and qualification logic are documented at the stated disclosure level. The standard is versioned so that each qualification can be interpreted against the exact methodology in force at the time of analysis.
Thresholds are checked against the source methodology for internal consistency, and literature-dependent methods are checked against their original publications. Independent reproduction can therefore be performed from the documented protocol and the candidate-specific qualification record.
