RBX1 measurement reliability: full-collection response and identifiability audit

Which conclusions about RBX1 measurement reliability are supported by the released experimental curves, and where do missing metadata, repeated-measurement variation or poorly identified models prevent a defensible affinity estimate? Produce a new, reproducible analysis of the entire published measurement collection. This is an evidence-quality investigation, not a nomination of the strongest published binder or a reconstruction of unavailable original participant methods.

Reconstruct every released measurement

Read rbx1.csv as UTF-8 with a possible BOM. Preserve all 322 candidate records and all experimental evaluations. Candidate identity is CSV id; the evaluations column is a JSON array. type, metric, target, value and unit retain their source meaning. The curve URL appears in the experimental spr_kinetic_curves or bli_kinetic_curves evaluation's value. Join that exact URL to the appendix file and retain its assay label; do not infer instrument equivalence from a similar filename.

Create coverage.csv, one row per candidate, and evaluation_inventory.csv, one row per experimental evaluation including its original array index. Record curve links, response labels, all reported physical parameters, expression observations, missing information, original author and supplied designMethod. The original method PDFs, initial ranked submissions and novelty checks are outside this input collection and cannot be reconstructed from a brief designMethod string. They are not required deliverables.

Analyse all 946 JSON files and all 4,656 trace objects, not only files attached to published positives. Each top-level JSON key identifies a trace whose object contains raw.t, raw.y, concentration, control, aggregated, virtual and fit. The trace ID is the exact appendix filename followed by :: and the original top-level JSON key. An analysis group is one candidate ID and one linked curve-file URL; its group ID is the compact UTF-8 JSON array [candidate_id,appendix_filename]. All group and trace sorting uses Unicode code-point order. Non-curve evaluation rows retain their CSV candidate ID and original evaluation-array index; these do not imply a trace linkage. The appendix covers the complete retrieved collection of linked experimental curves; it does not assert that every experiment or all instrument metadata were published. Report candidates with no measurement explicitly. Preserve all repeated evaluations; their order alone is not a laboratory replicate or run identifier, and it does not prove that one scalar evaluation corresponds to one particular curve file.

Create trace_diagnostics.csv, one row per top-level trace object, with candidate, URL/file and trace key, assay label, flags, concentration as stored, supported units or their absence, point/finite-pair counts, duplicate/backward time counts, first/last time, response extrema, first/last finite response and time-normalised trapezoidal area. The area uses finite pairs in original order and is null if fewer than two valid points, non-increasing time or zero duration prevents its definition; explain the reason. Also compute early-versus-late drift as mean(last k finite responses) minus mean(first k finite responses), where n is the finite paired-observation count and k = max(1, ceil(0.1*n)). Preserve original observation order, permit overlap when n is small, use each window's arithmetic mean and report null when n = 0. This is a response difference, not a slope divided by time. These are descriptive definitions, not biological pass thresholds. Record any additional noise statistic's exact definition.

Separate observed response from inferred kinetics

The released JSON does not explicitly label time, response or concentration units. All retrieved trace control flags are false; a zero concentration, a flag or a similar shape must not be silently promoted into a verified reference-subtraction history. Audit the full collection for usable references and identify exactly what they can establish. Keep instrument-specific amplitude scales separate. Do not claim an absolute kon, koff or KD from unlabeled coordinates alone.

For every candidate/curve-file group with finite time/response pairs, independently fit a baseline/drift model and a concentration-dependent response model. State model equations, observed-coordinate meaning, parameter domains, objective, residual treatment and stopping conditions. Fit measured raw.y; published fit response arrays cannot be training observations, fitted answers or a substitute for independent optimisation.

Where the stored fit.association.t and fit.dissociation.t arrays provide phase hints, they may be used as publisher-derived timing annotations, but their status must be explicit and their fitted y values must remain excluded. Evaluate at least three distinct phase/offset scenarios justified by observed timing whenever phase assignment is inferred. Where there are no defensible phase annotations, use a descriptive concentration-linked model on recorded coordinates and assess the phase ambiguity. Conditional 1:1 fits may be reported separately in source-coordinate units when the assumptions are explicit; lack of physical calibration does not prevent numerical residual or concentration-consistency analysis.

Write model_results.csv with all attempted groups, contributing trace IDs, model equations, fit objective, residual summaries, numerical parameters with coordinate meanings, convergence and boundary status. Compare baseline/drift and concentration-dependent models in model_comparison.csv. For each group with at least three distinct positive stored concentrations, hold out each concentration once, retaining whole traces, refit on the remaining concentrations and report prediction error as sqrt(sum((measured_response - predicted_response)^2) / N), over all N finite observed response/time pairs in the held-out traces, in recorded response coordinates. Predict held-out traces without fitting any parameter or offset to their response values; state the training-only rule for any trace-specific nuisance parameter. Report N and a reason-coded null if no finite predictions are available. For groups with fewer than three distinct positive concentrations represented by finite observed pairs, explain the missing comparison and still compute the available trace-level descriptors. Uncertain physical phase assignment does not waive held-out testing of the descriptive recorded-coordinate model. If that model cannot produce finite predictions, retain each required held-out case with its numerical failure and missing-data reason rather than silently dropping the group. Do not use published binding labels to determine which groups receive analysis.

Quantify identifiability of fitted response/time-scale parameters using a profile of the objective or resampling that preserves within-trace dependence, within the computational resource bounds below; numerical parameter domains must be stated and justified from recorded coordinate ranges and the model, rather than asserted as physical bounds. A local optimiser's success code or a Hessian standard error alone is insufficient. Show whether a parameter is supported by the observed response or only by imposed bounds. If physical calibration or phase interpretation is absent, the corresponding physical parameter is null with the exact reason; maintain separate conditional coordinate-parameter estimates and their sensitivity. Never turn an unbounded interval or nonconvergence into a precise affinity.

Measure consistency without inventing replicate linkage

Create candidate_consistency.csv for all 322 candidates. Within each assay, compare the available file-level response descriptors, concentration trends and conditional model parameters. Retain file identity and distinguish repeated reported measurements from independently established biological replicates. For each candidate and exact assay label, perform leave-one-file-out analysis of a stated robust centre and range over every file with a defined value of the same measured descriptor. Retain missing descriptor values and reasons; with fewer than two defined file values, state that file-to-file variability is unmeasured. This is a conditional comparison of recorded coordinates, not proof of common calibration or biological replication. Separately inventory the source attributes relevant to stronger comparability claims: stored concentrations and observed time ranges, control/aggregated/virtual flags, phase annotations and explicitly available units. Contradictory units or unsupported physical calibration prevent a physical-parameter comparison, but do not waive the recorded-descriptor sensitivity calculation. State and justify any further conditional model-parameter comparison from these source attributes; never infer comparability merely from evaluation order. Keep expression failure, reported nonbinding, missing fit and independently observed uncertainty distinct.

Create published_comparison.csv relating independently derived diagnostics to every published binding, expressed, binding_strength and kinetic-parameter evaluation. Preserve unknown scalar-to-file linkage instead of pairing values by array position. Compare distributions and candidate-level ranges where record-specific pairing is unavailable; do not manufacture one-to-one matches. Derive a collection-wide account of how often labels coincide with observable concentration dependence, inconsistent traces or unsupported kinetic precision. Quantify all denominators and expose individual contributing records.

For any candidate with both assay types, show their overlap and compare only justified, unit-aware quantities; if there are none, provide the empty overlap result. Cross-candidate SPR-versus-BLI differences cannot establish assay agreement. Include sensitivity to phase/offset scenarios, failed fits and treatment of repeated measurements; no favourable-only subset may determine the conclusion. A claim that a published label is unreliable requires quantified observed evidence and model limitations, not merely a disagreeing newly fitted number.

Provide figures covering collection-wide missingness, concentration dependence and held-out errors, identifiability and file-to-file disagreement. Also generate a diagnostic figure for every analysed curve file showing measured response, independent predictions where available and residuals. A useful conclusion may be that particular physical affinities cannot be identified from the release, but that conclusion must be backed by completed numerical analysis of the observable traces.

This buys a new analysis of historical published laboratory measurements. Attribute the observations to Adaptyv and the original authors where supplied. Do not claim new experiments, physical sample custody, original participant authorship, therapeutic effectiveness or a newly validated binder. The Poster accepts the publisher-controlled files below as publication provenance. A Solver-created hash is not independent laboratory authentication.

Use every listed measurement file and preserve the complete candidate population, failed measurements and missing observations. Published binding labels, affinities, fitted response arrays and graphics are comparison references, not correct answers that the analysis must reproduce. Do not use their response values to generate the independent fits. Never infer a successful experiment from a prediction, absent row or empty field. The task does not require a biological sequence or a sequence-design method.

Submit one ZIP, at most 40 MiB compressed and 250 MiB uncompressed, containing:

• report.md, at most 8,000 words excluding tables: question, methods and equations, evidence limitations, quantified results, diagnostic interpretation and conclusions linked to output rows and figures.
• run.py and readable supporting Python source: the full analysis, including all processing, fits, uncertainty calculations and figures. No pretrained model or undisclosed executable is allowed.
• analysis_plan.json: model families, formulas, parameter domains, phase/window rules, resampling unit and seeds, optimisation limits, exclusions and analysis scenarios actually used. This records reproducibility, not a claim of prospective preregistration.
• results/: the required UTF-8 CSV tables and PNG figures, plus checks.json containing coverage totals, numerical-validation results and run information.
• README.md: how the input filenames map to the appendix, explanation of output columns and the exact software versions used.

Do not duplicate the source datasets in the submitted archive. Both roles independently fetch the appendix inputs before execution, verify SHA-256, and mount them read-only under /inputs with the appendix filenames, preserving raw/ directories. Source ZIP files remain ZIP files; the analysis must read them without extracting outside its scratch directory. Python files run only in a fresh isolated Linux sandbox, without network, credentials, host directories or a container-control socket. Required runtime: Python 3.12, NumPy 2.1.3, SciPy 1.14.1, pandas 2.2.3 and Matplotlib 3.9.2; standard-library modules are allowed. No further package installation, paid service, language-model call or GPU is required or allowed during reproduction.

From the extracted submission root run python run.py --inputs /inputs --output /output --seed 20260909. The sandbox has four CPU cores, 8 GiB RAM and 10 GiB writable scratch space. One complete reproduction must finish within two hours and write all required outputs to a previously empty /output. Use exactly 100 repetitions whenever resampling is chosen, at most 10 starting points per fit and at most 2,000 objective evaluations per starting point; record every stopping reason, including nonconvergence. Use at least three dispersed initialisations for each nonlinear fit, unless an analytic or convex solution establishes the optimum. Run each initialisation to its declared numerical convergence criterion or the evaluation ceiling. A deliberately truncated search cannot establish scientific non-identifiability: distinguish numerical search failure from a supported flat or unbounded objective profile and report them separately. A budget-exhausted fit must remain visible and cannot be presented as an identified parameter. These are computational limits, not scientific thresholds. Never trade full-collection coverage for a selected set of published successes.

The Guardian runs one full reproduction per Submission. A second execution is allowed only after a demonstrated infrastructure fault; the Guardian must not silently fix Solver code. Unavailable inputs, decryption or sandbox infrastructure are operational blockers. Once those are available, a reproducible Solver-code failure or exhaustion of the stated resource cap fails the execution criterion.

Every result must be regenerated from the verified measurements. The Guardian compares the submitted and reproduced tables: row identities, labels, missingness and counts must agree exactly; finite numeric entries must satisfy abs(a-b) <= 1e-10 + 1e-6 * max(abs(a),abs(b)). This is a numerical reproducibility tolerance, not evidence that a scientific estimate is accurate. Reported conclusions must follow the calculations and their limitations; rounding for presentation must not change machine-readable results.

A Submission passes only if all of the following hold:

1. Input coverage and provenance: the full appendix is verified, every original candidate has a coverage row, every released raw trace has a trace row; non-curve experimental evaluations have inventory rows and do not require fabricated trace mappings, duplicate/unmapped records remain visible, and candidate linkage can be followed to the original source. A claim that a field is absent must be checked against the complete relevant source, not a selected example.
2. New computation: the complete numerical analyses required below are executed from the measurement arrays. Copied publisher tables, precomputed fitted arrays, a label filter, a narrative alone or a ranking of published KD values cannot satisfy the task.
3. Statistical validity within the stated claim: model formulas, fitting objectives, constraints, uncertainty methods, data reuse and exclusion reasons are explicit. Correlated time points are not counted as independent experimental replicates. Distinct assays and conditions are not pooled without an explicit justified comparison. Missingness, nonconvergence, boundary estimates and non-identifiability propagate to downstream conclusions.
4. Evidence-limited interpretation: units and phases are supported by the listed source or explicitly conditional. An unsupported physical quantity is null with a specific reason and supporting diagnostic, while computable descriptive statistics remain numerical. A blanket declaration that all data are unusable does not replace the required observable curve analysis. No universal confidence, binding or quality threshold may be invented and described as a laboratory rule.
5. Full reproduction: the prescribed command finishes within bounds and regenerates the required outputs within the numerical tolerance. The Guardian independently implements the elementary trace diagnostics for the first 12 trace IDs in ascending Unicode code-point order, or all traces if fewer than 12 exist, and checks them against the submission. The Guardian also traces source-derived analysis groups, ordered by the group IDs defined below: select the first, lower-middle and last group, with zero-based indices 0, floor((N-1)/2) and N-1; remove duplicate indices if fewer than three groups exist. Inspect every required scenario for each selected group from raw response through independent fitted prediction and residual calculation; when it has no physical estimate, inspect its conditional/descriptive model and the evidence for that limitation. With no usable group, verify the complete source-derived unusability ledger rather than inventing a model result. This audit supplements the full reproduction.
6. Complete scientific conclusion: the report answers the stated question using collection-wide results and uncertainty, includes the required sensitivity analyses, distinguishes disagreements from missing measurements, and explains what the available measurements cannot establish. The Guardian checks each claim against its cited generated table or figure; no external search or unpublished judgment rule is required.

All criteria are pass/fail. This bounty does not rank scientific truth by an invented numerical merit score. Among Submissions that pass every criterion, choose the numerically smallest on-chain Solver address, interpreted as an unsigned 160-bit hexadecimal integer. Only active Submissions count; Elgora permits one active Submission per Solver address and a replacement supersedes the earlier one. This identifies one winning active Submission. When every required artifact and input is available and no Submission passes, use no_valid_submission; zero active Submissions also uses that outcome. Unavailable evidence or evaluation infrastructure means no Verdict until the operational blocker is resolved, not automatic disqualification or permission to select another winner.

Forged measurements, concealed exclusions, fabricated provenance, mislabeled units, claimed independent experiments that did not occur, absent required outputs after successful artifact access, or code that requests access beyond the sandbox fail the relevant criteria. Do not include secrets, instructions to disregard this page or material unrelated to this analysis. Public Verdicts give criterion decisions and public-safe reasoning without publishing private Solver code, tables or report content. Guardian and settlement behavior otherwise follows Elgora's existing protocol.

Both Solvers and Guardians obtain the following files by unauthenticated HTTPS GET. Their listed bytes and SHA-256 define the accepted snapshot. If a URL serves different bytes, do not substitute its new content. No nested linked resource is required unless separately listed here. Filename mappings are part of this page; downloading does not execute any file. The source collection publishes observed sequences as historical identifiers; this bounty requires no sequence modification or design.

The CSV comes from the designated Proteinbase collection maintained for the GEM–Adaptyv results. Each curve URL must occur in that exact CSV and its downloaded bytes must match this appendix. Filenames beginning with raw/ are local names for the listed public JSON URLs; they are not additional hosted endpoints. The experimental publisher is Adaptyv; Proteinbase is its chosen publication source for this collection.