Mutation by Natural Dominance | Maldonado Research

Read this first

Research status

PR 49 · 5 November 2026. Learned detection thresholds, eligibility and route-classification uncertainty now share one declared error budget for conditional curvature inference. Full-cohort denominators retain gate failures and unresolved observations. Independent internal checks and synthetic replays verify the implementation under its assumptions. Methods 0.7.0 archives these methods; native biological calibration and validation of TMD remain unestablished. All 26 study and 18 planning inputs remain unresolved.

Earlier R43 · 16 October 2026. Authenticated barcode-processing sources specify an identity and annotation contract. An original structural preflight checks supplied keys and row roles and keeps unmatched annotations unknown. Eight adversarial tests and 67 separate internal assertions pass; structural success does not authenticate biological identities or the supplied row bindings. No actual CSV inputs or individual barcode-to-CNV assignments were recovered. All 26 study and 18 planning values remain NULL; native matched panels and biological fits remain zero. Methods 0.6.0 and the paused runner remain unchanged.

Earlier R42 · 16 October 2026. An authenticated author notebook preserves a summary totaling 80,751 grouped rows. Those rows are not verified as distinct biological lineages. The inspected Figure 7 recipe specifies 25 bins, correcting earlier KDE wording. A separate BASIL assessment distinguishes fitness inference from recovery of mutation births. Actual competition CSVs remain unrecovered; no native biological fit or new algorithm adoption is admitted. All 26 study and 18 planning values remain NULL; methods 0.6.0 is unchanged.

Earlier R41 · 15 October 2026. Exact conditional bounds show how repeated clone means can change a distribution when rows are counted instead of unique eligible clones. The calculation uses standard mathematics and hypothetical configurations; actual study distortion remains unknown. A separate April 2026 yeast growth–mutation study adds comparator context, without confirming TMD. The original fitness CSVs remain unrecovered. All 26 study and 18 planning values remain NULL; methods 0.6.0 is unchanged.

Earlier R40 · 15 October 2026. Authenticated source artifacts clarify how clone competition measurements and model-classified barcode lineages enter a fitness comparison. Clone/run multiplicity, time-unit calibration and barcode row mapping require the actual data. Official metadata identify three fitness CSVs, but their bytes remain unrecovered. No native Wsp/Aws/Mws panel or biological fit is admitted.

Earlier R39 · 15 October 2026. A 2022 yeast copy-number study compares inferred fitness effects with barcode and competition measurements, while empirical confirmation of its inferred mutation formation rate remains open. Its population trajectories reuse the 2018 experiments; later assays add measurements. A static source-code audit identifies positive-coefficient filters and unresolved data joins. No native Wsp/Aws/Mws panel or biological fit is admitted.

Earlier R38 · 15 October 2026. A published yeast copy-number reporter study provides an implemented example of reporter controls, selected-clone DNA checks and separate fitness measurements. Its 7.1% population detection threshold, fluorescence purity correction and later molecular checks have different measurement targets. The reviewed assessment specifies what independent genotype, ancestry and negative-control records a native mutation-origin test still needs. No native panel, biological fit or new theorem is established.

Earlier R36 · 14 October 2026. A newly inspected calibration supplement explains the 60,000-read subsampling used for Figure S3 and provides selected read-processing settings. A separate native SBW25 preprint adds genotype, cell-state and founding-density context for culture-mat outcomes. Those measurements remain distinct from mutation births and single-new-mutant establishment. Actual dose, preparation ancestry, candidate exclusion records and a complete native panel remain unresolved; biological TMD validation remains pending.

Earlier R35 · 14 October 2026. Official archive records now identify the experiments and libraries behind all 22 replayed calibration rows. Two additional V4 libraries labeled 300 mg are candidates for the condition absent from the processed table. A shared registered BioSample does not establish physical preparation ancestry or independent biological trials. Actual dose uncertainty and per-run exclusion counts remain unavailable; a complete native Wsp/Aws/Mws panel and biological TMD validation remain pending.

Earlier R34 · 14 October 2026. Original published data and recipe archives now support a verified source-calibration replay. An original Python calculation reproduces all six Figure 2D slope/R² values from 22 supplied accession rows. Unknown replicate ancestry, absent attempted-condition records and post-lysis standard entry limit biological interpretation. A complete native Wsp/Aws/Mws panel and biological TMD validation remain pending.

Earlier R33 · 14 October 2026. Two published methods provide implemented DNA-standard calibration resources, and an actual public data/code deposit identifies a concrete next source archive. Their post-extraction or post-lysis entry stages determine which measurements can be calibrated. Earlier cell losses and mutation births still require separate evidence. Source-data replay, a complete native Wsp/Aws/Mws panel and biological TMD validation remain pending.

Earlier R32 · 13 October 2026. The recent HiDenSeq preprint clarifies a non-UV fitted calibration for its UV comparison and explicitly leaves general mechanism identifiability open. A separately fitted calibration needs joint uncertainty and a justified frozen transfer. These distribution comparisons do not establish a unique biological cause. Supporting information and underlying data remain unacquired; a complete native Wsp/Aws/Mws panel and biological TMD fit remain unestablished.

Earlier R31 · 13 October 2026. The source authors explain why their fitness assays need a declared competitor and frequency context; the original fitness aggregation remains unresolved. A thesis adds historical reconstruction detail without identifying the present MPB stocks or an independent replication. A recent HiDenSeq preprint offers a barcoded measurement lead, with pooled-lineage, calibration and yeast-to-WS transport limits. A complete matched Wsp/Aws/Mws panel and biological TMD fit remain unestablished.

Earlier R30 · 13 October 2026. A later primary study reports the projected mwsR mutation in newly evolved SBW25-WS3, without proving identity with historical WSR or MPB10613. The inspected linked code models pathway activity; it does not explain the fitness summaries. Reweighting the three visible coefficients cannot match the stored mean and standard deviation under ordinary population-SD assumptions. Other uncertainty definitions and corrections remain unresolved. A complete matched Wsp/Aws/Mws panel and biological TMD fit remain unestablished.

Earlier R29 · 13 October 2026. Exact reference-sequence slices reveal a CDS-origin difference that changes how the same mwsR genomic variant is named. They also support the historical awsX 33-base interval, while the newer stock annotation remains unresolved. A conditional calculation excludes a single missing equal-weight coefficient as an explanation of the disputed fitness summary. Actual stock identity and the intended aggregation still require evidence. A complete matched Wsp/Aws/Mws panel and biological TMD fit remain unestablished.

Earlier R28 · 11 October 2026. The historical WS collection now has a traceable common-ancestor sampling frame and an explicit molecular classifier, including one unresolved isolate. Two exact publisher workbooks expose mutation and competition measurements, with one fitness summary requiring reconciliation. A practical review worksheet specifies the evidence needed for the next study decision. A complete matched Wsp/Aws/Mws panel and biological TMD fit remain unestablished.

Earlier R27 · 11 October 2026. The author supplement identifies reported Wsp, Aws and Mws alleles and strain IDs, and distinguishes microscopy media and strain backgrounds. Two further small licensed workbooks expose date-block and plate-level measurements. The resulting control requirements separate mutation supply, establishment and later recovery. A complete matched Wsp/Aws/Mws panel and biological TMD fit remain unestablished.

Earlier R26 · 8 October 2026. Numerical PTATO accuracy and mutation-origin benchmarking are parked while independent truth and complete inputs remain unresolved. Two small licensed workbooks from a bacterial adaptation study are authenticated. Their colony and surface-count records sharpen the next control requirements, with no complete matched Wsp/Aws/Mws panel or biological TMD fit admitted.

Earlier R25 · 8 October 2026. The official PTATO sources now yield a 477-folder topology, seven scoped inventories covering 31 files, and three exact small callability summaries. This is not a complete recursive file manifest or a replay of variant and locus recovery. Large inputs remain unacquired. A conditional code audit clarifies which rows enter the comparisons; its constructed examples are not study results.

Earlier R24 · 8 October 2026. The official PTATO dataset license and exact Table S1 and S2 sample metadata are now authenticated. Sample relationships, training flags and a WT display-label mismatch can be checked; variant and callability inputs still need a bounded replay. No numerical variant or mutation-origin benchmark is admitted. All 18 native-control planning fields remain unresolved.

Earlier R23 assessment. Exploratory research. A review of official PTATO validation materials verified software releases and inspected figure-analysis code. Reported detection rates depend on selected variants, sample quality and the DNA regions that can be evaluated. Complete input data and authentic sample metadata were unavailable, so no numerical variant or mutation-origin benchmark is admitted. Correctly identifying a variant does not establish when it arose.

All 26 actual bacterial study inputs remain unresolved. Qualified biological and statistical review is pending; biological validation and cancer prevention have not been established. Methods 0.6.0 is archived separately: it freezes the 1 October methods snapshot and excludes the later GitHub research rounds.

What the public work reports

One error budget connects learned gates and observation bounds

A predeclared plan allocates uncertainty across threshold calibration and future count bounds, then propagates eligibility and classification limits into simultaneous curvature intervals. The calculation retains every cohort unit, including unresolved outcomes. Its guarantees require independent truth, suitable sampling, frozen score pipelines and justified transfer from controls to native observations. Synthetic checks establish computational behavior; they supply no native error rate or biological validation.

R43 barcode identity contract and structural preflight

Newly authenticated author scripts extend the observation contract from barcode extraction to retained rows and annotation joins. The original table preflight requires explicit keys and row roles, rejects ambiguous structure, and keeps unmatched annotations unknown. Eight adversarial tests and 67 separate internal assertions pass on synthetic examples. Supplied identities and roles remain unverified biological and source-row claims; no actual CSV inputs or barcode-to-CNV assignments were recovered. Read the source contract ↗ Read the preflight requirements and limits ↗

R42 rendered outputs and binned-plot correction

The authenticated Figure 7 notebook stores four grouped counts totaling 80,751 rows and requests a binned ridgeline plot with 25 bins. Input identifiers, unique biological lineages and publication execution remain unverified. This corrects the earlier KDE description. A separate BASIL source assessment explains why fitness estimates for detected lineages do not recover all mutation births or invisible losses. Read the rendered-output audit ↗ Read the separate BASIL assessment ↗

Repeated clone means permit exact conditional weighting bounds

If the inspected pipeline gives each eligible clone one common mean repeated one to three times, standard mathematics bounds the difference between a distribution over rows and one counting each clone once. Exact finite-count bounds, constructive examples and tests make the assumptions inspectable. The examples are hypothetical; actual multiplicities and publication execution remain unknown, so no published bias is established. Read the conditional bounds and their limits ↗

Fitness comparisons depend on clone weights, time units and barcode identity

An authenticated source audit distinguishes estimates from clone competition measurements from model-classified barcode lineages. The competition script can retain repeated clone means across fitted runs, while the barcode pipeline requires a mapping through filtering and any pooled low-count row. Actual weighting and unit calibration remain unverified. Official metadata identify the three fitness CSVs; their bytes remain unrecovered. Read the fitness pipeline and source-record audit ↗

Fitness validation leaves CNV formation rates unconfirmed

Avecilla et al. (2022) reuse the 2018 evolved populations and compare inferred beneficial fitness effects with barcode estimates and clone competitions. The paper calls for empirical confirmation of the inferred formation rate. An authenticated source-code audit documents positive-coefficient filters and unresolved input joins. The clone workbook and required CSV records were not recovered in this audit, so their rows were not replayed or reconciled. Read the inference and validation assessment ↗

Reporter validation needs a defined observation target

A distinct assessment of Lauer et al. (2018) separates a population detection threshold, regrown-clone fluorescence checks, selected late-clone sequencing and pairwise fitness. These controls support different source-specific claims; they do not identify every earlier mutation birth or single-new-mutant establishment. Official deposit metadata identifies named next artifacts, while file contents and reuse terms remain unverified. Read the measurement-admission assessment ↗

Processing scope and native cell state clarify the next test

All 22 Figure S3 source rows total 60,000 reads, consistent with the newly inspected supplement, while the earlier replay uses different Figure 2D counts. A separate native preprint measures historical WS genotypes and reports genotype-dependent reporter activity and density-sensitive mat persistence. Protocol filters, gated cells and intact culture mats do not supply native mutation-event truth or independent recovery calibration. Read the combined source assessment and remaining records ↗

Named archive records narrow the missing-condition question

All 22 retained rows now have exact run, experiment and library joins to one registered BioSample. Two released V4/300-mg libraries are named candidates for the omitted condition, but their processed standard counts and exclusion decisions are unverified. Six low-mass rows carry 37-mg versus 37.5-mg labels across source figures; the original replay is preserved. These records clarify provenance without supplying biological independence or the full attempted-sample denominator. Read the identifier audit and remaining uncertainty ↗

Published calibration numbers are reproducible

The authenticated Figure 2D processed table reproduces the archived author output under original Python and independent high-precision arithmetic. The three descriptive slopes are about 0.876, 0.868 and 0.900, with R² about 0.991, 0.974 and 0.953. These are retained-source read-ratio fits; the accession rows do not establish independent biological trials or complete failed-sample counts. Read the numerical replay and its limits ↗

Implemented standards clarify the measurement stage

Barcoded genomic-DNA standards enter the portable BSINC method after extraction. Synthetic rDNA standards can enter soil processing after cell lysis. These published methods calibrate downstream DNA quantities under their tested conditions; they require separate evidence for earlier cell recovery, viability and mutation origin. A public source-data deposit supplies a specific next artifact before numerical replay. Read the implemented calibration resources and limits ↗

Calibration and competing mechanisms need separate checks

The HiDenSeq authors fixed the UV comparison’s reads-to-mutants conversion using a non-UV model fit and report that inferred induced rates agree with independent mutant colony counts before and after UV. Known-abundance spike-ins are a future proposal. The authors also leave open whether mechanisms such as mutant fitness and phenotypic switching can be distinguished from distribution shape alone. These details strengthen the requirements for independent calibration, justified transport and complete alternative models in a TMD test. The preprint’s supporting information and underlying data remain unacquired. Read the calibration and mechanism assessment ↗

Fitness needs its competitor and frequency context

In the inspected author response, Lind and colleagues explain that WS fitness depends on frequencies and was measured against a wspF WS competitor rather than the wild-type ancestor. Their final model does not combine mutation rates and selection coefficients to predict the evolution experiments. Those fitness summaries need context before reuse; reviewer suggestions and reporting-template questions cannot supply a native calibration. Read the reporting-record and author-response assessment ↗

A recent barcode method suggests a measurement route

The HiDenSeq preprint posted on 7 August 2026 uses barcodes to sample fluctuation-distribution shape in yeast. Its pooled barcode counts are lineage labels, rather than automatically independent mutation events, and its authors identify competition as a confound requiring calibration. The method is a lead for feasibility review; native Wsp/Aws/Mws units, supply and event truth remain unadmitted. The inspected version has not been certified by peer review. Read the recent method lead and transfer limits ↗

The same allele does not establish the same stock

The later study identifies newly evolved SBW25-WS3 with reported mwsR genomic position 5856845 and V728A. This corroborates the earlier coordinate projection, but a matching focal mutation cannot link that isolate to historical WSR or MPB10613. Physical stock identity still needs pedigree, background and sequence records tied to the actual material. Read the primary stock-context assessment ↗

Separate simulation code from the fitness calculation record

The named supplement and linked repository describe pathway-activity simulations. Their inspected contents do not supply the disputed fitness aggregation. A standard conditional bound also excludes reweighting the three visible coefficients alone: at the stored mean, their smallest population SD is about 0.02107, above the stored 0.01515. SEM, transformations, control corrections or different observations remain possible explanations to investigate; author values are retained. Read the source-record and conditional calculation assessment ↗

Pin the CDS coordinate origin before comparing an allele

The returned reference annotation places the mwsR coding start 102 bases after the frame implied by the 2019 mutation workbook. Under those respective frames, the same genomic change is named V728A and V694A. Gene-relative labels therefore need a declared sequence and annotation frame. This reference check does not authenticate an experimental stock. Read the coordinate and source-join assessment ↗

Check an explanation against both reported summaries

The fitness workbook carries the disputed mean and standard deviation into its embedded chart. Retaining its three visible coefficients, a fourth equally weighted value cannot produce both stored summaries under an arithmetic mean and sample standard deviation. The intended calculation remains unknown and the author values are retained. Read the conditional reconciliation assessment ↗

Declare which outcome and classifier are being tested

The older collection sampled one WS isolate after five days from each of 26 common-ancestor microcosms. Its molecular table assigns 16 Wsp, 6 Aws and 3 Mws isolates, leaving one unresolved. Counting that isolate as functionally Wsp-associated gives a different classification. These sampled outcomes do not identify every mutation birth or the first arriving lineage. Read the source and classifier assessment ↗

Make the next measurement decision concrete

The review packet maps 44 unresolved study and planning fields to specific evidence questions, with eight ordered gates and an unpriced resource worksheet. It asks whether establishment can be observed before terminal recovery and whether independent ancestry and lost-lineage truth are available. This preparation supports a qualified feasibility decision; no experiment or external review has occurred. Read the practical review packet ↗

Experimental context travels with the strain

The reported wspF, awsX and mwsR alleles now have source strain IDs. Some microscopy uses LB medium and a pyoverdin-related deletion, whereas the default experiments use King’s B medium. Those measurements require an explicit background and context match before they can support a common comparison. Read the strain and measurement map ↗

Separate establishment from later detection

At one nominal dilution, separate mat-success date blocks report 0/3, 1/3 and 9/9. This variation does not identify a cause, but dilution alone cannot define a matched context. The reporter workbook records 3,376 sampled colonies across six plate labels; these categories do not identify molecular mutation routes. Independent lineage outcomes before recovery, followed by recovery and classification controls, are needed for the proposed comparison. Read the phase-specific control requirements ↗

A benchmark needs independent truth and complete inputs

The scoped PTATO calculation needs 18 data objects, with more than 1.64 GB of known source-file sizes and further unresolved inputs. Running the whole figure script requires at least 34 dataset objects plus runtime resources. A larger download would supply computational inputs, while independent genotype and origin truth would still need separate evidence. Read the concrete replay and truth decision ↗

Ecological context belongs in the establishment controls

A recent SBW25 study observes ancestral cells facilitating mutant colonization. Exact licensed workbooks contain sampled-colony categories and early surface counts for wspF, awsX and mwsR genotypes. The next controls need to match density, resident ancestral scaffold and timing. Colours, morphologies and dimensional density indices do not supply independent mutation origins or establishment probabilities. Read the source and measurement audit ↗

Small callability summaries can be checked exactly

Two WT summaries and one MSH2 summary match their official file sizes and hashes. Their frequencies use a shared denominator that excludes bases labelled REF_N; including the reported REF_N frequency makes the total exceed one. This is a verified arithmetic convention, not a demonstrated source-result error. The summaries do not establish which genomic positions or variants pass the comparison. Read the scoped inventory and summary audit ↗

Which rows count changes the comparison

The archived code uses variant eligibility pooled across samples, which can differ from eligibility within each sample. Averaging only present categories can also differ from including zero-count categories. Constructed examples demonstrate these distinctions; their actual effect on the study results remains unknown. The sample-specific subset uses PTATO’s own labels and is not independent truth. Read the conditional denominator audit ↗

Authenticated metadata clarifies the validation frame

Official Table S1 (79 rows) and Table S2 (16 rows) match their catalogued file sizes and hashes. The four-group recovery frame includes a FANCC branch marked as training material. A sample marked Training=No does not by itself establish donor-level holdout. Two WT labels are reversed between the sample table and recovery-bar code; this is an annotation mismatch, with no numerical-result error demonstrated. Read the metadata audit and remaining admission gates ↗

First establishment can differ from eventual fixation

An unreviewed Park–Krug preprint posted 7 October 2026 describes a lineage that establishes first yet loses to a later competitor. Conventional clonal interference can therefore affect the declared endpoint. Its extension to more alternatives supplies a least-fit marginal probability; a complete three-route forecast remains missing. No parameter transfer, validated TMD comparator or simulation replay is claimed. Read the source assessment and model limits ↗

Variant accuracy differs from origin accuracy

A correct inherited variant can still be wrongly assigned as a new mutation origin. The control plan therefore asks for distinct evidence on genotype errors, ancestry and acquisition episodes, missing lineages and transfer into the actual biological system. Published coverage, conditional classifier sensitivity and agreement between assays support different claims. The packet is ready for qualified assessment; no external review or experiment has been obtained.

Support and reference rules can hide true events

In a synthetic model, two descendants each have a one-half chance of a correct call. A rule requiring both can have inclusion probability one-half, one-quarter or zero, depending on their joint calls. Requiring a reference call adds another dependence condition. Sharp small-model bounds have exact attaining examples and independent certificates. These rules illustrate a measurement problem; they do not reconstruct a native caller, infer unseen births or validate cancer prevention.

A reviewable cell-lineage study contract

The new pedigree ledger distinguishes observed divisions, inferred origin intervals, inherited copies and unknown genotypes. A clearly labelled synthetic example keeps dead and lost lineages visible and separates branch-supported hypotheses from ambiguous leaf observations. The expert-review packet asks qualified biologists and statisticians to assess actual event definitions, native caller rules, controls, replication and inference before a pilot. No laboratory protocol or experiment has been commissioned.

Equal cell recovery can give unequal event detection

In a synthetic validated-call model, two descendants each have a one-half chance of providing a true call. Their shared mutation can have detection probability one-half, three-quarters or one, depending on joint recovery. Repeated inherited copies remain one birth. Exact bounds and classical fixed-event sampling calculations make these assumptions explicit; they do not reveal unknown births or fit a biological mutation rate. The any-call rule is separate from a caller requiring multiple subclone support.

Observed pedigrees offer a concrete measurement lead

A published epithelial cell-line experiment combines optical division tracking with sequencing of related descendants. Eleven of 45 available HT115 channel cells and 13 of 26 RPE1 cells produced primary sequence samples. The 24 subclones descend from two displayed founders; these realized nested fractions are not calibrated route capture or independent mutation births. Lost genotypes and unmatched cell backgrounds keep this study outside a confirmatory TMD test.

Bounded uncertainty separates unresolved patterns from incompatible ones

Under an explicitly admitted expected-mass model and positive independent Cartesian parameter bounds, exact curvature intervals show whether some common value remains possible. Wide overlap is inconclusive; empty intersection excludes the restricted model only under those population assumptions. A separate proportionality check asks whether ordinary supply, growth and recovery explain all three routes. Passing the necessary curvature condition alone is weaker. No biological confidence interval, count law, power or prevention effect is supplied.

A cancer-relevant test needs independent lineage evidence

Two foundational studies make strong competing explanations concrete. A human normal-esophagus study maps 844 spatial samples from nine donors; detectable cancer-associated clones are distinct from malignant transformation or counted new mutation births. A mouse study tests competition between normal epithelial clones and early lesions, with observations, interventions and simulations kept separate. Its authors leave analogous human tumor elimination unknown. These established results guide a proposed measurement contract; they do not validate TMD.

Ordinary growth or recovery can produce the same endpoint pattern

Exact synthetic examples keep mutation supply fixed while changing relative terminal yield or detection. The supply-adjusted curvature remains common across contexts. This concerns normalized expected endpoint composition, not a complete lineage, count or first-arrival law. A useful test needs independently measured alternatives and predictions fixed before the held-out outcomes. Cancer routes and their ordered axis are still unchosen.

Reporting scale matters before biological inference

A declared clinical supplement remains unacquired after four distinct retrieval attempts. A conditional synthetic example has a raw midpoint median of 50 and a back-transformed log midpoint of 10 after zero handling. Another separates a ratio of marginal medians from a paired effect. These examples preserve the published numbers and do not authenticate the paper's actual reporting convention or calibrate biological recovery.

Cells, colonies and recovery need separate denominators

A bounded source screen examines two primary main-text methods/results studies: processing of 72 clinical urine specimens and persistence assays in P. aeruginosa PA14 and clinical isolates. Twenty measurement-ledger rows separate specimens, flow events, regrowth and CFUs. Membrane-integrity counts and colony formation are different outcomes; a nonsignificant comparison does not establish equivalence. Reported marginal CFU medians are not paired recovery probabilities. Neither source calibrates capture for SBW25 or the three TMD routes. NanoSpacer remains an abstract-only lead.

What known absolute capture could identify

Under a finite compound-Poisson model, known independent per-unit capture, a justified bound on strictly positive terminal marks and the complete population jump-intensity vector permit triangular inversion of the positive mark weights and intensity. This is ideal conditional identification, not an estimator from a finite culture histogram. Sharp void-probability bounds are deterministic model bounds, not observed confidence intervals. The producer passes 1,624 exact synthetic checks; independent internal review passes 10,747 invariant checks and 20,154 producer comparisons across 960 synthetic cases. These are computational checks, not biological observations. No biological capture, mutation rate or new theorem is established.

The remaining invisible events and unit ambiguity

Unknown capture can leave different event intensities with the same complete observed law. Even at known capture, extra precursor events with zero terminal descendants remain invisible. Identification of positive terminal intensity does not recover those lost events or prove that a CFU represents one cell. Ancestry, imported alleles, growth, recovery and native/reporter transfer still need matched measurements. No actual study field or biological calibration is supplied.

Authenticate estimator inputs before fitting

In the same released SBW25 source data, nine of twelve wild-type culture rows yield fractional genotype-corrected candidate-pool estimates. Candidate colonies, confirmed colonies, CFU densities and nominal whole-culture totals remain different quantities. Their reconstruction does not authenticate what was submitted to FALCOR or establish a software error. The historical input report, software build and estimator settings remain unverified; no biological mutation-rate fit is justified.

A paired check of the observation process

Under independently justified disjoint assignment from one culture, allocation between two samples conditional on their total follows a binomial split law. Shared-culture disjoint samples, overlapping recapture and independent cultures can share marginals while differing jointly. Exact finite examples pass 8,208 producer checks and 8,200 independently implemented review checks. Those numbers count computational assertions, not biological trials. No paired biological data, rate fit or confidence interval is supplied.

Passing a technical split check has limits

Many terminal histories can pass the conditional allocation check, including imported alleles, common unmeasured loss and sampling of multicell CFU packets. Passing does not uniquely authenticate molecular origin, absolute recovery, individual-cell units or the sampling frame. An explicit model allowing precursor events with zero terminal descendants changes latent event intensity while preserving the full descendant-count distribution. This is a conditional counterexample, not a general claim that every strictly positive clone model is unidentifiable.

Nominal plate geometry and conditional detection

The SBW25 source recipe and workbook reconstruct sampling fractions for all 24 culture occurrences: 0.0075%–0.075% for the reporter background and 7.5% for the deletion background. These are nominal aliquot fractions, not calibrated detection probabilities. A classical clone-and-sampling benchmark shows why zero observations alone can confound mutation intensity with clone survival and capture. Full count distributions can add information under a justified model. All new numerical examples are conditional and synthetic; no biological mutation-rate fit, new theorem or outward numerical certificate is claimed.

An inherited allele can enter without a new mutation

The E.coli–P1 paper behind the earlier abstract lead has now been read at full-main-text depth. Its reported roughly 380-fold excess of recovered recombinant over mutant CFUs at 24 hours concerns descendant abundance, not independently counted molecular event rates. Imported donor alleles, newly created mutations and standing variation have different origins and denominators. The source supports an assay-specific transfer mechanism; it does not establish transfer in SBW25. Supplementary models and raw data were not acquired or reproduced.

Public reporter measurements, reproduced

The licensed figure-data archive for Farr and colleagues' SBW25 hotspot study is now acquired and checksum verified. Original code reconstructs 24 culture occurrences and passes 986 source arithmetic, formula and mapping checks. The pooled corrected terminal-frequency ratio between reporter backgrounds is about 118.97. The authors' reported MSS mutation-rate point estimates instead imply 58.33. These are different quantities: mutant descendants and observation affect terminal frequencies. The mutation-rate fit and its intervals were not reproduced.

Keep the assay and sampling gaps visible

The transcript reconstruction gives a 5.9603 geometric-mean ratio of regional transcript proxies under the supplied ideal-doubling transform; it does not calibrate native mutation supply. One culture row lists seven counted candidates but eight sequenced colonies, and the first sequencing occasion has 94 available reads against 95 workbook entries. These are open sampling and provenance questions. Ten measurement-contract gaps remain; colonies, sequence reads and repeated transformants cannot be relabelled as independent Wsp/Aws/Mws trials.

What a supply-transfer assumption would change

An ordinary algebraic calculation shows how prespecified native/reporter supply-error bounds would widen the existing route-contrast intervals together with recovery uncertainty. The dataset supplies no finite native/reporter bound for the three routes. Without such a justified bound, the calculation supplies no biological test. It establishes neither a new theorem nor an additional causal dominance mechanism.

Certified comparisons of frozen forecasts

A new engine gives rigorously bounded score comparisons for prespecified forecasts of observed qualifying Wsp/Aws/Mws outcomes. One simultaneous confidence region supports every frozen comparator. Twenty synthetic tests and 30,887 independent exact review assertions passed; the stored scoring example is inconclusive. These checks verify the bounded numerical procedure; independent founders, justified count laws and predictions fixed before outcomes still need a real study contract.

A precision limit, before outcomes

A conservative illustration targeting a 0.05-nat interval half-width requires 527 and 921 qualifying outcomes in its two contexts, exceeding the prototype's 500-unit per-context cap. This is a bound on interval width, not a power result or a recommended biological sample size.

New source checks strengthen competing explanations

Full-text inspection of Farr and colleagues (2025), Barber and Couce (2026), and Li and Gyorgy (2025) examines transcription-dependent mutation supply, population history and plasmid masking. Farr's engineered SBW25 reporter shows about a 58-fold difference in reported fluctuation-rate estimates across the compared backgrounds; that does not calibrate native Wsp/Aws/Mws supply. The review separates mutation production from establishment and phenotypic visibility. The Farr figure-data archive is now audited separately above; the Barber/Couce and Li/Gyorgy datasets remain unacquired. No complete matched Wsp/Aws/Mws panel is admitted within the inspected scope.

Recovery uncertainty, including all-success controls

Methods extension 0.5.0 adds conservative confidence bounds for recovery-corrected Wsp/Aws/Mws route comparisons. All-success controls still leave uncertainty; observed zeros can produce unbounded ranges. The package includes 13 automated tests and 160 synthetic studies. No complete matched biological route/control panel has been authenticated within the inspected material.

Correcting the source record

The Wsp/Aws/Mws counts 46/41/18 in Lind et al. (2019) came from separate pathway-isolated strains. They cannot be treated as competing-route winner frequencies.

Context in published observations

A secondary analysis restores the three experimental blocks in the published rpoB table. Medium-specific predictions outperform a pooled prediction in each held-out block, with an exploratory total gain of 13.126 bits across 111 isolates. This does not identify the proposed TMD mechanism.

A test of changing route fractions

A context-stratified timing diagnostic accounts for right censoring. Selected synthetic null scenarios rejected 12/200 and 9/200 datasets at a 5% threshold; a deliberately strong switching alternative was detected in 200/200. These are simulations of a procedure, not biological experiments.

A limit on identifying mechanisms

The same Lomax first-event waiting-time distribution and constant route shares can arise from different rate constructions. First-event observations alone cannot distinguish those mechanisms.

Latest GitHub research checkpoint

PR 49: integrated threshold and observation inference. The public methods and conditional proof connect maximum-selected thresholds, full-cohort eligibility, wrong-label limits and curvature inference under one prespecified error budget. The fixed plan is separate from later observations; gate failures and unresolved outcomes remain in the denominator. The interface does not validate biological truth, chronology, independence or transport merely by checking declarations and hashes.

Independent internal review, finite-case calculations and scratch replays check the conditional implementation. All bundled observations are synthetic; the work supplies no native calibration, measured biological effect, validated TMD mechanism or new theorem. Methods 0.7.0 archives the connected observation and control methods with their separate contracts and replay instructions.

Earlier R43: barcode identity contract and structural preflight. The reviewed checkpoint and source report separate extraction, retention, aggregation, numerical export, ordered identifier binding and CNV annotation. Equal array lengths cannot authenticate a positional pairing, and missing annotation does not establish a negative genotype. Saved notebook counts selected by different predicates lack a shared run identity; their differences do not establish attrition or author error.

The original structural preflight requires explicitly keyed analyst-prepared tables and declared row roles. It rejects duplicate identities and positional-only inputs, retains zero estimates, preserves unmatched annotations as unknown, and accounts for declared aggregate/padding rows and orphan annotations. The validation contract records eight adversarial tests; separate internal checking adds 67 assertions. These synthetic table checks authenticate neither the supplied key/role binding nor genotype, time calibration, execution or native admission. No source-author code or biological simulation was run. Actual CSV bytes and individual barcode-to-CNV assignments remain unavailable. All 26 study and 18 planning inputs remain NULL. Predictors, comparators, evaluation, error allocation, fixed stopping, the paused runner and methods 0.6.0 remain unchanged.

Earlier R42: rendered outputs and binned-plot correction. The reviewed checkpoint authenticates the original-author Figure 7 notebook and its stored summary. The 80,751 total counts grouped rows after identifier binding and a CNV annotation join; it does not authenticate unique biological lineages, negative genotypes or mutation births. The source audit records the category definitions, typed column reports and source comparisons. Visible and saved R source agree on the statistical recipe while several input-directory assignments differ. Actual CSV bytes and the publication execution remain unverified.

The inspected plot call specifies binline with 25 bins, correcting earlier KDE and automatic-bandwidth wording. The prior conditional CDF bounds do not directly certify this plot or apply to the barcode summary. A separate assessment of BASIL (2026) distinguishes inference of fitness from mutation-birth and missing-event truth. It does not adopt BASIL as a TMD comparator. All 26 study and 18 planning inputs remain NULL. Predictors, comparators, evaluation, error allocation, fixed stopping, the paused runner and methods 0.6.0 remain unchanged.

Earlier R41: conditional clone bounds and a separate recent source. The reviewed checkpoint derives exact bounds for the selected Lauer-derived component when the unchanged source pipeline retains each finite clone mean one, two or three times. A CDF records the fraction at or below a chosen value. The largest possible row-to-clone CDF gap has supremum 2 − √3, approximately 0.267949, over unknown population compositions; no finite integer population attains that value exactly. Known row and clone counts can tighten the bounds. These are standard bookkeeping results, not a new theorem, confidence interval or measured effect in the publication. The assumption contract requires stable clone identity, finite common means and the same ancestor and sign selectors; the appended barcoded component is outside this application.

A separate assessment of Ding et al. (29 April 2026) examines growth and mutation-rate measurements in budding yeast. Its growth proxy, generation denominator, mutation classes and detection rules inform conventional comparators. It does not supply a native Wsp/Aws/Mws panel or confirm TMD. Actual competition CSVs remain unrecovered; all 26 study and 18 planning inputs remain NULL. Predictors, comparators, evaluation, error allocation, stopping, the paused runner and methods 0.6.0 remain unchanged.

Earlier R40: fitness pipeline and source records. The reviewed source assessment authenticates competition and barcode processing artifacts, including the author’s bc04-to-bc01 alias statement and MIT software license. Repeated clone means can affect distribution weights if this pipeline generated the consumed table. Barcode fitness values can inherit a model likelihood decision; they are not independent observations of mutation births. Actual CSV contents, row mapping, time-unit calibration and exact publication execution remain unverified. All 26 study and 18 planning values remain NULL. No numerical replay, native biological fit or new theorem is admitted.

Earlier R39: fitness validation and mutation-rate inference. The reviewed source and code assessment distinguishes experimental fitness checks from model-dependent formation-rate estimates in an existing yeast cohort. The beneficial-variant comparison applies a positive-coefficient filter; it does not establish all-clone membership or a DNA-validated negative set. Required CSV inputs and the original clone workbook were not recovered. All 26 study and 18 planning values remain NULL. No source numerical replay, native biological fit or new theorem is admitted.

Earlier R38: reporter validation and origin requirements. The reviewed source assessment and 12-gate matrix identify the independent molecular, ancestral and loss-control records needed before a reporter-labelled lineage can support a native-origin claim. The older yeast source is new to the inspected TMD ledgers, not a newly published discovery. All 26 study and 18 planning values remain unresolved. No source numerical replay, native biological fit or new theorem is admitted. R37 unchanged access failures remain parked.

Earlier R36: source stages and native alternatives. The reviewed assessment verifies a source-table subsampling distinction and inspects native expression, reporter and culture-mat readouts. The native genotype family is historical; four author-reported RNA culture replicates per genotype do not become new mutation origins. Original PDF access and stated dataset-link destinations remain unresolved. No source pipeline, new regression, native biological fit or new theorem is admitted. All 26 study and 18 planning values remain unresolved.

Earlier R35: official sample-provenance audit. The identifier assessment verifies all 22 retained run joins against official metadata and names SRR30280226 and SRR30280227 as candidate omitted V4/300-mg libraries. The 24 released condition labels do not establish 24 attempted biological samples. No processed exclusion counts, physical replicate genealogy or actual input-dose uncertainty are recovered. All 26 study and 18 planning values remain unresolved.

Earlier R34: verified processed-source replay. The calibration replay authenticates two original archives and joins the supplied table, recipe and archived output. All 22 supplied rows are retained without new exclusions, pseudocounts or weighting. The V4 300-mg condition is absent from the processed table; input-dose uncertainty and replicate provenance remain unresolved. Source arithmetic is reproduced; no native biological fit is admitted. All 26 study and 18 planning values remain unresolved.

Earlier R33: implemented calibration resources. The resource assessment compares two published methods, their standard materials, entry stages and measured units. Digital-PCR agreement on the same extracted DNA does not validate prior extraction losses; rRNA-gene quantities require assumptions before conversion to cells. The named data archive remains unacquired. A scoped local filename check identifies no new eligible biological artifact. All 26 study and 18 planning fields remain unresolved; model, forecasts, eta, error allocation and stopping are unchanged.

Earlier R32: calibration and mechanism identifiability. The methods assessment distinguishes separately fitted calibration from reuse of evaluation endpoints, and selected-model discrimination from unique mechanism identification. The named supplement request returned a rate limit; an indexed main-text representation supplied the new context. A PMC metadata record describes the same preprint, and an exact stock search identified no new stock record within two excerpts. All 26 study and 18 planning fields remain unresolved; frozen forecasts, eta, error allocations and stopping are unchanged.

Earlier R31: source context and a measurement lead. The reporting-record assessment distinguishes author limitations from reviewer suggestions and leaves Figure 9 aggregation unresolved. The historical stock search adds reconstruction context from the same earlier work without a current MPB identity record. The recent method lead specifies barcode-unit, calibration and native-transport questions for qualified review. All 26 actual study inputs and 18 native-control planning fields remain unresolved; model, forecasts, eta, error allocations and stopping are unchanged.

Earlier R30: stock context and calculation records. The stock assessment distinguishes a newly evolved isolate from unauthenticated historical and MPB stock joins. The calculation-record assessment identifies the purpose of linked simulation code and constrains a weights-only explanation under explicit SD assumptions. All 26 actual study inputs and 18 native-control planning fields remain unresolved; model, forecasts, eta, error allocations and stopping are unchanged.

Earlier R29: coordinate frames and calculation constraints. The source-join assessment records exact reference slices, gene aliases and a conditional crosswalk between coding frames, without selecting or authenticating actual study stocks. The summary assessment checks the embedded chart and constrains the missing-coefficient explanation under explicit assumptions. All 26 actual study inputs and 18 native-control planning fields remain unresolved; model, forecasts, eta, error allocations and stopping are unchanged.

Earlier R28: source identity and practical feasibility. The source assessment retains the historical sampling frame, molecular-versus-functional classification, allele annotation questions and an unreconciled workbook summary. The qualified-review packet requests independently supported units, predictors and claim-specific origin controls. All 26 actual study inputs and 18 native-control planning fields remain unresolved; model, forecasts, eta, error allocations and stopping are unchanged.

Earlier R27: strain identity and measurement contracts. The source map records the reported alleles, modified backgrounds, date blocks and unresolved culture/well/plate joins. The control assessment specifies the independently measured denominators required to distinguish supply, establishment and recovery, and requires conventional forecasts to predict the same observed W/A/M categories. All 26 actual study inputs and 18 native-control planning fields remain unresolved; model, forecasts, eta, error allocations and stopping are unchanged.

Earlier R26: input decisions and bacterial measurements. The benchmark input/truth assessment makes the scope and cost of a PTATO replay explicit. The licensed bacterial source assessment preserves a summed total of 1,014 sampled colonies, including 21 in the separate ambiguous Dark wheel category, missing cells and genotype-labelled surface units. These are source arithmetic, not mutation-birth counts or a biological fit. All 26 actual study inputs and 18 native-control planning fields remain unresolved; model, forecasts, eta, error allocations and stopping are unchanged.

Earlier R25: source identity and comparison rules. The catalogue and summary audit authenticates three small source summaries without acquiring the large variant or callable-locus inputs. The code and denominator audit keeps global and sample-specific eligibility, means with and without zero categories, and pooled fractions distinct. Neither result supplies an independent genotype or mutation-origin benchmark. All 26 actual study inputs and 18 native-control planning fields remain unresolved; the model, forecasts, eta, error allocations and stopping rule are unchanged.

Earlier R24: authenticated processed-data metadata. The official PTATO metadata audit verifies dataset-specific CC BY 4.0 terms, a four-file root inventory and exact Table S1/S2 bytes. This strengthens sample-to-clone/culture mapping and documents training overlap and the WT display-label reversal. It does not complete the recursive data manifest, reproduce variant recovery or identify independent mutation origins. The next step is a bounded inventory and replay of paired variant and locus-callability inputs, with the sample mapping and all exclusions retained.

Earlier R24: endpoint and model scope. The 7 October mutation-bias preprint assessment distinguishes first establishment from ultimate fixation under continued mutation and competition. The least-fit marginal in its multi-alternative extension cannot fill the complete Wsp/Aws/Mws forecast. The source is an unreviewed preprint; parser output was assessed without a numerical simulation replay. This adds a concrete question about competing lineages to the existing design, with all 26 study inputs and 18 native-control planning fields unresolved and forecasts, eta, error allocations and stopping unchanged.

The earlier R23 review inspects official PTATO software and analysis artifacts, identifies the study’s raw-data accession and processed-data version, and checks the conditions behind reported variant recovery. The analysis code derives candidate timing from earlier sample calls and needs a data-level check of how repeated variants enter the eligible comparison. Independent ancestry, acquisition-episode truth and lost-lineage measurements remain missing. Read the artifact review and admission limits ↗

Separate hypothetical control calculations illustrate how zero observed errors and positive prevalence affect error bounds and predictive value. They do not estimate native error rates, establish power or choose a study size.

The preceding checkpoint adds a native-control evidence plan and exact synthetic calculations for support and reference detection rules. Three primary-source assessments separate callable coverage, conditional classifier performance, event-origin truth and assay concordance; one assessment is restricted to selected paragraphs. Blind mathematical comparisons agree with exact certified bounds across 2,889 marginal cases. No actual native origin calibration, qualified external review or biological TMD test is claimed. The earlier checkpoints summarized below retain their original scope; the raw supplement reporting map remains blocked.

The recovery-transport follow-up combines the earlier sensitivity analysis with certified confidence bounds. It reuses all 160 published synthetic studies and reports 44 new passing tests. The chosen recovery bounds are hypothetical sensitivity settings, rather than measured biological recovery errors.

Two subsequent source-audit rounds are complete. The first checks census measurements and the distinction between evolutionary histories, cultures and technical plate counts; it does not estimate biological drift or establishment for TMD. The second authenticates four primary articles and records sampling units, reused observations and missing measurements. No complete matched Wsp/Aws/Mws panel is admitted within that inspected scope. This is an eligibility result, not a test rejecting or confirming TMD.

The third round completes a prospective prediction design as a design-only candidate. It specifies how to reserve an entire context, freeze predictions before selected outcomes, and compare independently measured mutation supply, establishment and recovery explanations. Its forecast concerns observed qualifying Wsp/Aws/Mws outcomes; biological interpretation additionally requires justified recovery controls and their transfer to the experiment. Twenty-six required study inputs remain unresolved, including genotypes, sampling frames, contexts, sample sizes, stopping rules and decision thresholds. All design checks use synthetic inputs; no biological experiment or registered study is claimed.

The latest dataset audit preserves the culture/transformant/occasion hierarchy, nested colony confirmations and missing or unresolved sequencing outcomes. It strengthens a concrete competing mutation-supply explanation, while leaving native-site supply, establishment and recovery transfer unresolved. No actual study-registry input is filled from the selected reporter endpoints.

The original October watch screened two PubMed abstracts and metadata. The Euglena mutation-accumulation record remains inspected at that depth. A subsequent source assessment reads the complete main E.coli–P1 paper, including methods and embedded captions; supplementary files and raw measurements remain unacquired. This additional reading of the same study is not an independent biological replication, and no parameter is transferred to SBW25.

The fluctuation follow-up resolves nominal plate-volume bookkeeping while leaving biological capture, growth, clone fitness and confirmation assumptions open. Its classical model and bounded synthetic checks are separate from the earlier certified-score engine. The prepared request for original likelihood inputs and sample mapping remains unsent; no actual study field is resolved.

The inheritance follow-up distinguishes spontaneous mutation from import of an existing donor allele. A mutation-only entry baseline needs ecological evidence or an explicitly measured origin-inclusive domain. Recovered colonies and endpoint allele identity do not separately identify molecular origin, event intensity, establishment or recovery.

The source-input follow-up reuses the same authenticated 24-culture dataset and separates exact reconstructed vectors from the missing historical software submission. Bounded software/source-access attempts acquired no authoritative run artifact. The source-fields request remains unsent; original raw inputs and earlier audit receipts are preserved.

The paired-aliquot follow-up is a conditional observation-quality design candidate based on established marking and compound-Poisson identities. No actual allocation ratio, rejection rule or study protocol has been frozen. It separates physical disjointness, overlapping recapture and independently justified culture histories. A passing allocation gate cannot fill the biological registry. The next priority is matched cell/CFU and absolute capture/recovery calibration with ancestry, founder and missingness records.

The recovery-source assessment screened eight PubMed metadata/abstract candidates and inspected two selected primary main-text methods/results sources. Its 20-row ledger preserves exact observation units and unresolved normalization, missing-data and source-to-native transfer. It supplies no numerical capture parameter for SBW25; the newest microscopy lead remains abstract-only.

The capture-calibration package uses finite exact synthetic models to distinguish an identifiable positive-mark case from unknown-capture and invisible-zero-mark ambiguities. Its full population vector is assumed, not recovered from observed finite counts. All 26 actual study inputs remain unresolved. The next priority is cell/CFU calibration validation with matched lineage controls, independent introduction and capture denominators, and explicit recovery and ancestry records.

Read the earlier R23 GitHub research checkpoint and its limits ↗

What natural dominance means here

TMD describes a route as more naturally dominant in a specified context when it has a higher rate of producing a defined successful heritable outcome. Mutation supply, accessibility and establishment can all contribute. This differs from genetic dominance in a heterozygote. Observed winner frequencies alone cannot separate these causes, selection, drift or recovery bias. The empirical question is whether a prespecified cross-context restriction predicts held-out observations after independently measured supply and justified recovery correction. A fitted route preference alone does not establish a new causal mechanism.

Assumptions and open questions

The timing diagnostic requires independent units, independent censoring conditional on context, exact first successful arrival times and time bins chosen before inspecting outcomes. Endpoint counts in the rpoB table do not satisfy those timing requirements. Failure to reject constant route shares does not prove TMD. The newer confidence method also requires justified count laws, introduced control totals, prespecified sampling and evidence that relative control recovery transfers to biological samples. Overlapping intervals do not establish equivalence or identify a mechanism. Certified forecast scores additionally require a complete prespecified endpoint ledger and justified conditional count laws, with Other, no-qualified-outcome, unresolved and missing outcomes retained. Zero qualifying outcomes are unevaluable; correlated descendants or forecasts fitted after outcomes do not satisfy the current contract.

Read the source materials

Repository documentation is the primary source for this overview. Research links are pinned to their public GitHub checkpoints; archive links below identify published versions. The recovery-source ledger, capture-calibration candidate, earlier dataset and source assessments, paired sampling design, certified-score implementation and prospective design are separate from the labelled 0.5.0 archive and the frozen 0.6.0 archive.

Archives and citation

Methods release 0.7.0, DOI 10.5281/zenodo.23970140, is published in the existing report/methods concept family 10.5281/zenodo.23068055. The methods snapshot is dated 4 November 2026 (America/Los_Angeles), with source and computational audit dated 5 November UTC. It archives conditional observation and control calculations, synthetic examples and bounded replay checks, including PR 49. Related methods retain their own assumptions and are not automatically combined. The archive establishes neither native biological validation nor a registered experiment.

Historical methods release 0.6.0, Zenodo record 23219776, is published in report/methods concept DOI family 10.5281/zenodo.23068055. Its immutable package freezes the 1 October methods snapshot; later source audits, designs and calculations remain outside that archive. The archive does not add biological validation or include the latest PTATO review. Read the verified publication-state note for the archived files and scope.

The earlier labelled 0.5.0 release is record 23075312. Historical unversioned record 23113326 retains the inherited 0.5.0 file inventory. The expanded report 0.1.0 remains record 23068056, and historical software v2.6.5, record 22398093, belongs to a separate publication family. Cite the version actually used and attribute the original data papers.

Overview updated from the merged PR 49 integrated methods and published methods 0.7.0 release guide and verification evidence on 4 November 2026 (America/Los_Angeles; source and computational audit 5 November UTC). Reviews reported in the source materials are internal computational and source checks, rather than external peer review. Consult the cited source version for full statements, qualifications and reuse terms. Author: Ricardo Maldonado.