Build the physical layer before claiming the biology.

Program Status

The Genome Physics Initiative is at the foundational benchmark and implementation-contract stage.

No Genome Physics module is commercially available today.

The initiative begins with a bounded scientific task:

Given a short DNA or RNA system, its declared physical state, environment, and optional perturbation, calculate local geometry, pairing, stacking, mechanics, opening behavior, and physical deltas from Flux physics.

Only after that layer survives frozen, held-out, blind, and prospective tests should the program advance into:

• long-sequence scanning;
• mutation mechanism;
• oligo and guide triage;
• therapeutic redesign;
• protein–nucleic-acid interaction;
• regulatory sequence physics;
• multiscale genome context.

The program is ambitious because it aims to create a new physical computation layer for biological sequence.

It is disciplined because every higher claim depends on lower-layer evidence.

1. Why the Initiative Exists

Modern genomics can read sequence at enormous scale.

It can identify:

• genes;
• motifs;
• variants;
• expression patterns;
• statistical associations;
• evolutionary conservation;
• likely functional regions.

These capabilities have transformed biology.

They do not fully answer a different question:

What did the sequence change physically?

A mutation may alter:

• base pairing;
• stacking;
• local geometry;
• bend and torsion;
• opening behavior;
• groove presentation;
• hydration and ion interaction;
• protein or ligand engagement;
• the set of accessible structural states.

An oligo may fail because it:

• folds onto itself;
• dimerizes;
• presents the wrong geometry;
• binds weakly;
• binds too broadly;
• loses mismatch discrimination;
• depends on an inaccessible target state;
• becomes sensitive to a resistance mutation.

A construct may create physical risk through:

• stable hairpins;
• repeats;
• local strain;
• alternative structures;
• self-binding;
• mutation-sensitive regions;
• manufacturing-sensitive states.

These are physical questions before they become biological outcomes.

The Genome Physics Initiative exists to build the missing bridge:

Sequence
    ↓
Physical state
    ↓
Structure and mechanics
    ↓
Molecular interaction
    ↓
Mutation or design delta
    ↓
Experiment

The program does not begin by predicting phenotype from text.

It begins by calculating the physical state the sequence creates.

2. Existing Foundations

FluxMateria is not beginning with a blank software stack.

The existing Matter Computing Platform already provides adjacent foundations in:

• molecular representation;
• atomic and bond properties;
• geometry and conformation;
• environmental calculation;
• protein and docking workflows;
• mutation and candidate enumeration;
• high-throughput search;
• provenance;
• frozen prediction;
• benchmark and validation infrastructure.

These foundations do not prove Genome Physics.

They reduce the amount of infrastructure that must be invented before the scientific work can begin.

The initiative must still derive and validate new nucleic-acid physics, including:

• nucleotide state;
• pairing;
• stacking;
• backbone geometry;
• sequence context;
• mismatch behavior;
• environmental dependence;
• structural alternatives;
• RNA-specific states;
• protein–nucleic-acid interaction;
• multiscale biological closure.

The strategic advantage is therefore not that the problem is already solved.

It is that the platform already contains the machinery required to turn a successful derivation into:

• code;
• search;
• provenance;
• a user interface;
• a validation program;
• a commercial workflow.

3. Program Principles

3.1 Physical property before biological label

The first outputs must be:

• geometry;
• energy;
• mechanics;
• structural state;
• molecular interaction;
• perturbation delta.

The first outputs must not be:

• pathogenicity;
• expression;
• clinical significance;
• editing outcome;
• cell behavior;
• therapeutic efficacy.

Higher-layer hypotheses may be generated later, but they must remain explicitly separate from the directly calculated physical result.

3.2 State before score

A sequence string does not fully define a physical system.

Every calculation must declare, where relevant:

• DNA or RNA chemistry;
• strand count and orientation;
• pairing state;
• modification state;
• temperature;
• salt and ions;
• pH;
• solvent or hydration;
• topology;
• molecular partners;
• force, confinement, or boundary conditions.

The scientific rule is:

Sequence identifies the matter. State determines the physical problem.

3.3 Local before global

The program begins with bounded local windows and short structures.

It does not begin with:

• atomistic chromosomes;
• whole-cell models;
• complete gene regulation;
• whole-genome phenotype.

Local physics is:

• testable;
• parallel;
• cacheable;
• easier to falsify;
• directly reusable in later layers.

3.4 Prediction before optimization

The engine must first predict held-out physical measurements.

Only then may it:

• rank candidates;
• redesign sequences;
• optimize constructs;
• compile therapeutic objectives.

3.5 Emergence before adjustment

Every authoritative Genome Physics property must emerge from Flux physics.

The property may not be determined or repaired by:

• empirical nearest-neighbor tables;
• fitted melting formulas;
• fitted salt corrections;
• empirical mismatch penalties;
• motif property lookup;
• learned folding energies;
• learned accessibility;
• learned off-target scores;
• assay-trained residuals;
• benchmark-tuned thresholds.

Those systems may remain:

• external baselines;
• comparators;
• candidate generators;
• structural proposal tools;
• search priors.

They may not carry physical evidence.

3.6 Blind evidence before availability

A module does not become available because:

• the interface is complete;
• the API runs;
• the runtime is fast;
• a retrospective result looks promising;

Availability requires the relevant evidence gate.

3.7 AI supports search and execution

AI may:

• generate candidate sequences;
• propose structural states;
• prioritize calculation;
• write and test software;
• organize literature;
• summarize Decision Objects.

AI may not:

• determine the authoritative property;
• correct a residual;
• substitute a learned physical output;
• promote a candidate without Flux recomputation.

3.8 Safety before unrestricted design

Sequence design must remain bounded to legitimate scientific and therapeutic objectives.

High-risk biological design requires:

• explicit purpose;
• target and organism context;
• human review;
• audit logging;
• safety screening;
• refusal or escalation where appropriate.

4. The First Scientific Contract

The first release is deliberately small.

Required inputs

• DNA or RNA sequence;
• nucleic-acid chemistry;
• strand count and direction;
• pairing state;
• supported modifications;
• temperature;
• salt and relevant ions;
• pH or protonation assumption where required;
• solvent or hydration state;
• reference and altered state for physical deltas.

First authoritative outputs

• base-pair geometry;
• dinucleotide-step geometry;
• twist and rise;
• roll and tilt;
• shift and slide where supported;
• bend and torsional response;
• pairing contribution;
• stacking contribution;
• local opening or separation tendency;
• mismatch or mutation delta;
• environmental sensitivity;
• provenance and applicability.

Excluded first-release claims

• pathogenicity;
• gene expression;
• complete cellular accessibility;
• chromatin state;
• editing efficiency;
• delivery;
• immune response;
• clinical effect.

The first scientific contract is successful only if the same frozen physical implementation performs across unseen sequence contexts and declared environmental conditions.

5. The Physical Claim Ladder

Every output must be classified.

P0 — Represented state

Examples:

• sequence;
• strand state;
• environment;
• modification;
• supplied structure.

P1 — Direct physical property

Examples:

• geometry;
• pairing;
• stacking;
• mechanics;
• opening behavior.

P2 — Interaction or perturbation delta

Examples:

• mismatch delta;
• mutation delta;
• oligo–target interaction delta;
• reference-versus-mutant target delta.

P3 — Mechanism hypothesis

Example:

The mutation may weaken target presentation through increased local opening and changed protein-contact geometry.

P4 — Biological hypothesis

Example:

The mutation may reduce cleavage or expression under the declared biological context.

P5 — Therapeutic or clinical outcome

Example:

The candidate may improve treatment response.

The authority boundary is:

P0–P2
    Direct Genome Physics scope after validation

P3
    Mechanistic interpretation requiring supporting evidence

P4–P5
    Additional biological, pharmacological, experimental,
    regulatory, and clinical evidence required

The program must never allow a P1 or P2 result to be displayed as P4 or P5 merely because the explanation sounds plausible.

6. The Planned Module Family

The architecture contains eight modules.

None is currently available.

1. Genome Physics Scanner — planned first release

Intended decision:

Where are the local physical hotspots, unusual states, and mutation-sensitive regions?

Planned outputs:

• position-by-position physical tracks;
• local geometry;
• mechanics;
• opening tendency;
• perturbation sensitivity;
• structural-state warnings.

2. Variant Mechanism — planned first release

Intended decision:

What did the variant change physically?

Planned outputs:

• reference-versus-altered physical delta;
• affected structure or interaction;
• mechanism hypothesis;
• decisive experiment.

3. Oligo Triage — planned first release

Intended decision:

Which small candidate set should be synthesized first?

Planned outputs:

• target interaction;
• self-structure;
• mismatch discrimination;
• off-target physical compatibility;
• mutation robustness;
• primary and backup families.

4. Oligo Designer — planned second release

Intended decision:

How should a weak candidate be redesigned?

Planned outputs:

• constrained variants;
• property trade-offs;
• Pareto-ranked redesigns;
• preserved-function report.

5. Resistance Rescue — planned first release

Intended decision:

Which compounds or sequence designs retain interaction across a mutation panel?

Planned outputs:

• reference-versus-mutant target delta;
• lost and retained interactions;
• candidate ranking;
• rescue hypotheses;
• assay shortlist.

6. Construct Optimizer — planned second release

Intended decision:

Which construct regions create physical risk, and how may they be redesigned while preserving function?

Planned outputs:

• structural-risk map;
• self-binding and alternative-state hotspots;
• preservation-aware redesign;
• manufacturing hypotheses.

7. Protein–Nucleic-Acid Physics — planned later release

Intended decision:

How does a protein, nuclease, polymerase, or other molecular machine engage the declared nucleic-acid state?

8. Regulatory Designer — planned later release

Intended decision:

How should a bounded regulatory sequence be physically presented under declared constraints?

The following are foundations rather than separate modules:

• Nucleotide and Duplex Core;
• Structural-State Search;
• RNA Physics Core;
• Environment and Modification Layer.

The following remain research programs:

• chromatin and chromosome-scale physics;
• Genome Compiler.

7. Foundational Workstreams

Workstream A — Nucleotide and Duplex Core

Objective:

Build the first authoritative physical layer.

Required work:

• nucleotide representation;
• backbone state;
• canonical pairing;
• sequence-dependent stacking;
• local geometry;
• mechanics;
• mismatch;
• modifications;
• temperature and ion dependence;
• state and unit schema.

Promotion gate:

Pass the foundational Geometry packet and the required portions of Thermodynamics, Mismatch, Mechanics, and Environment.

Workstream B — Structural-State Search

Objective:

Generate and evaluate physically plausible local and short structures.

Required work:

• hairpins;
• stems;
• loops;
• self-dimers;
• alternative duplex alignments;
• short RNA states;
• candidate-state deduplication;
• Flux evaluation of promoted states.

Promotion gate:

Demonstrate that candidate generation remains non-authoritative and that Flux-calculated state ranking generalizes to hidden structures.

Workstream C — Genome Physics Scanner

Objective:

Compile the local kernel across long sequences.

Required work:

• sliding windows;
• motif caching;
• boundary handling;
• variant overlays;
• construct scanning;
• provenance;
• applicability warnings;
• deterministic scaling tests.

Promotion gate:

Validated local tracks plus measured long-sequence execution correctness and runtime.

Workstream D — Mutation and Variant Physics

Objective:

Calculate physically meaningful reference-to-altered deltas.

Required work:

• substitutions;
• mismatches;
• short insertions and deletions where represented;
• damage and modifications;
• sequence-context transfer;
• interaction deltas.

Promotion gate:

Held-out perturbation classes and sign-correct physical deltas.

Workstream E — Oligo and Guide Physics

Objective:

Reduce candidate spaces before synthesis.

Required work:

• candidate enumeration;
• self-structure;
• target interaction;
• mismatch discrimination;
• off-target physical compatibility;
• chemical and manufacturing constraints;
• backup families.

Promotion gate:

Prospective top-k enrichment relative to declared baseline selection.

Workstream F — Therapeutic Redesign and Rescue

Objective:

Move from analysis to bounded inverse design.

Required work:

• oligo redesign;
• mutation-aware compound or sequence rescue;
• construct redesign;
• preservation constraints;
• Pareto ranking;
• prospective experiment.

Promotion gate:

A generated candidate improves a predefined experimental endpoint.

Workstream G — Interaction and Regulation

Objective:

Extend the local kernel into protein–nucleic-acid and bounded regulatory contexts.

Promotion gate:

Mode-specific physical evidence and a separate biological validation plan.

Workstream H — Multiscale Genome Context

Objective:

Compile local physics into nucleosome, chromatin, loop, and chromosome representations.

Status:

Research frontier.

Promotion requires validated lower layers and an explicit multiscale information-preservation contract.

8. Foundational Benchmark Program

The first benchmark architecture contains six packets.

Packet G — Geometry

Targets:

• base-pair geometry;
• dinucleotide-step geometry;
• twist;
• rise;
• roll;
• tilt;
• shift;
• slide;
• bend;
• width or groove-related quantities where defined.

Primary source candidates:

• Nucleic Acid Knowledgebase;
• RCSB Protein Data Bank.

NAKB is the successor to the Nucleic Acid Database and provides search, annotation, and downloads for experimentally determined nucleic-acid-containing structures.

Packet T — Thermodynamic and energetic behavior

Targets:

• pairing and stacking behavior;
• duplex or hairpin stability;
• ordering across sequences and motifs;
• opening or separation work;
• transfer across temperature.

Primary source candidates:

• Array Melt dataset;
• DNA calorimetric force-spectroscopy data;
• curated primary optical-melting and calorimetry studies.

The 2025 Array Melt study reports a published dataset of 27,732 sequences with two-state melting behavior. It is an evaluation source—not a source of Flux property values.

Packet M — Mismatch and perturbation delta

Targets:

• canonical-to-mismatch delta;
• geometry delta;
• energetic delta;
• sequence-context transfer;
• modification delta.

Sources should include primary studies with:

• explicit sequence;
• explicit conditions;
• direct measurements;
• reproducible inclusion rules.

Packet K — Mechanics

Targets:

• bending response;
• torsional response;
• persistence-related quantities;
• force–extension behavior;
• salt and temperature dependence.

Sources should prioritize single-molecule and direct mechanical experiments with downloadable curves or source data.

Packet E — Environment

Targets:

• temperature;
• salt and ion dependence;
• pH where supported;
• solvent and hydration;
• environmental deltas.

Environmental conditions may define the system.

Fitted environmental correction formulas may not determine the Flux property.

Packet R — Short RNA states

Targets:

• short stems;
• hairpins;
• bulges;
• mismatches;
• local geometry;
• state ranking.

Primary source candidates:

• RNA STRAND;
• NAKB and RCSB experimental RNA structures;
• primary optical-melting and structural studies.

RNA STRAND is a curated collection of experimentally determined or comparative-analysis-supported RNA secondary structures and is useful for proposal, evaluation, and split design.

9. Cohort and Leakage Rules

The benchmark must be frozen before hidden evaluation.

Required deduplication

• exact sequence;
• reverse complement where applicable;
• near-identical motif context;
• duplicated experimental measurements;
• homologous structural family;
• near-identical modification pattern;
• related structures from the same experimental series.

Required holdout axes

At least one hidden evaluation must hold out:

• sequence contexts;
• motif families;
• lengths;
• mismatch classes;
• modification types;
• environmental conditions;
• source laboratory where practical.

Evaluation isolation

Development teams may access:

• development data;
• published baselines;
• development metrics.

They may not access:

• hidden target values;
• hidden score reports before freeze;
• held-out annotations that reveal outcome.

Use separate:

• development manifest;
• validation manifest;
• hidden-test manifest;
• prediction archive;
• scorer.

Prediction freeze record

• source-data hashes;
• inclusion and exclusion rules;
• deduplication report;
• split manifest;
• baseline versions;
• formula and code commit;
• state schema;
• metrics;
• pass thresholds;
• prediction archive.

A benchmark is hidden only when the development team cannot inspect target values before the prediction freeze.

10. Metric and Promotion Policy

Metrics must match the property.

Possible metrics include:

• MAE;
• RMSE;
• angular error;
• physical-unit error;
• rank correlation;
• sign accuracy;
• top-k enrichment;
• interval coverage;
• applicability coverage;
• outlier rate;
• runtime and scaling.

Final pass thresholds must be frozen before hidden evaluation.

They should be tied to:

• experimental repeatability;
• simple sequence baselines;
• established empirical models;
• molecular-mechanics or high-cost methods;
• modern learned baselines;
• the intended commercial decision.

The pass policy is:

No post-result tuning.

If the primary metric fails:

• repair implementation;
• derive missing physics;
• improve state representation;
• or restrict the claim.

11. Development Gates

Gate 0 — Program Contract

Freeze:

• state schema;
• property definitions;
• authority rules;
• source record;
• benchmark manifests;
• safety boundary.

Gate 1 — Foundational Kernel

Deliver:

• nucleotide and duplex representation;
• local geometry;
• pairing and stacking;
• mechanics;
• mismatch deltas;
• environment handling.

Evidence:

• mathematical consistency;
• retrospective physical benchmark;
• held-out sequence-context test.

Gate 2 — Scanner

Deliver:

• long-sequence execution;
• position tracks;
• mutation overlays;
• provenance;
• applicability.

Evidence:

• local-track validation;
• long-sequence correctness;
• measured runtime.

Gate 3 — Triage

Deliver:

• oligo or guide candidate ranking;
• self-structure;
• target interaction;
• mismatch discrimination;
• off-target panel.

Evidence:

• blind or prospective top-k enrichment.

Gate 4 — Redesign

Deliver:

• constrained variants;
• preserved requirements;
• property trade-offs;
• experiment-ready shortlist.

Evidence:

• prospective design improvement.

Gate 5 — Interaction and Regulation

Deliver:

• protein–nucleic-acid or bounded regulatory workflows.

Evidence:

• mode-specific physical and biological tests.

Gate 6 — Multiscale Research

Deliver:

• coarse-grained nucleosome and chromatin representations.

Evidence:

• lower-layer validation;
• information-preservation audit;
• separate multiscale benchmarks.

Skipping a gate is not permitted.

12. First Commercial Validation Programs

Resistance Mutation and Rescue

Partner provides:

• reference target;
• mutation panel;
• candidate compounds or sequences;
• hidden assay results where possible.

FluxMateria freezes:

• target-state deltas;
• candidate ranking;
• lost and retained interactions;
• rescue hypotheses.

Partner measures:

• retained activity;
• enrichment;
• mechanism agreement.

Oligo or Guide Triage

Partner provides:

• target region;
• candidate pool or design constraints;
• chemistry and assay conditions.

FluxMateria freezes:

• self-structure;
• target interaction;
• mismatch discrimination;
• off-target physical compatibility;
• ranked primary and backup families.

Partner measures:

• activity;
• specificity;
• enrichment;
• synthesis burden avoided.

Construct Risk and Redesign

Partner provides:

• construct;
• preserved function constraints;
• manufacturing or stability assay.

FluxMateria freezes:

• structural-risk map;
• redesign candidates;
• expected direction of change.

Partner measures:

• predefined physical or manufacturing endpoint.

These programs are deliberately bounded.

They do not require the platform to claim complete biological prediction.

13. Evidence of Success

The initiative succeeds in stages.

Scientific success

A frozen physical kernel predicts unseen nucleic-acid structure, mechanics, mismatch, or thermodynamic behavior.

Search success

Top-ranked candidates are enriched for measured success relative to baseline selection.

Mechanism success

The platform identifies a physical reason for failure that is confirmed experimentally.

Design success

A generated candidate improves a predefined endpoint.

Operational success

A partner reduces:

• synthesis;
• assays;
• design cycles;
• time;
• uncertainty;
• cost.

Illustrative outcomes include:

• reducing 1,000 candidates to 20;
• enriching top-ranked oligos for activity;
• separating resistance mutations by physical impact;
• producing a better backup family;
• improving a construct endpoint through bounded redesign.

These are success definitions.

They are not current achievements.

14. Claims Require Evidence

The initiative will not claim:

• clinical diagnosis;
• universal pathogenicity;
• complete expression prediction;
• whole-cell behavior;
• whole-genome physical evidence;
• editing efficiency from duplex physics alone;
• therapeutic potency from target interaction alone;
• delivery or immune safety;
• autonomous organism design;
• unrestricted genome optimization.

It will not confuse:

• local opening with cellular accessibility;
• mutation delta with clinical effect;
• guide binding with editing outcome;
• structural risk with manufacturing performance;
• target interaction with in-vivo efficacy.

The physical layer is valuable precisely because its claim is bounded.

15. Safety and Biosecurity

The program is intended for legitimate:

• therapeutic research;
• mutation and resistance analysis;
• oligo, guide, primer, and probe design;
• bounded construct engineering;
• diagnostics and assay design;
• biosensor development;
• physical validation research.

High-risk sequence-design requests must be:

• restricted;
• escalated;
• logged;
• reviewed by a human;
• refused where appropriate.

The Genome Compiler remains a bounded research frontier.

Its work remains bounded to reviewed use classes, evidence gates, and biosecurity controls.

16. Conclusion

Genome Physics is the next frontier because biological sequence is matter organized into state, structure, interaction, and function.

The initiative will not begin by claiming all of biology.

It will begin by calculating what a sequence physically is.

Its scientific discipline is:

Build the physical layer before claiming the biology.

Its state rule is:

Sequence identifies the matter. State determines the physical problem.

Its development rule is:

Predict before optimizing. Local before global. Evidence before availability.

Its authority rule is:

Every authoritative property must emerge from Flux physics.

Its commercial purpose is:

Reduce the physical design space before expensive experiments begin.

Flux Theory opened the door.

Matter Computing made Genome Physics possible.

The Genome Physics Initiative is the program that will build it.

Appendix A — Program Gate Summary

Appendix B — Module Readiness

No Genome Physics module is commercially available today.