When the expected binding route is unavailable
In a BioTwin planning example, the absence of an accessible cysteine led to candidate plans spanning other residue chemistry, interface engagement, allostery and recruitment. The suggestions included prerequisites and unresolved questions.
- Change the chemistry · Investigate whether another residue and recognition mechanism could support a useful chemical interaction. Non-Cysteine Covalent Design
- Change the state or interface · Explore whether an allosteric effect or a different interaction surface could produce the desired response. Programmable Allostery
- Change how the protein is handled · Investigate a recruitment or degradation mechanism, with the required steps made explicit. E3 Recruiter and Molecular-Glue Mechanism Design
Drug discovery on the computer you already own.
BioTwin can run locally on regular desktop computers and laptops. A smaller team can investigate targets, compare molecular hypotheses and build on prior findings without first making a dedicated computing cluster the entry price for its research.
- Start with a focused question · Use structural, geometric and reduced-model analyses to decide which possibilities deserve closer investigation. Explore the methods
- Keep the investigation connected · Carry the structures, assumptions and evidence forward as you compare approaches, using BioTwin’s 8DB foundation. See the foundation
- Size the work to your machine · Memory, target size, sampling depth and the chosen method determine practical capacity. Larger or more detailed workloads may need additional compute. Discuss your workload
Spend compute where it can change the decision.
Test an access constraint directly. Ask another question of a prepared state model. Add detailed simulation or experiment where it could change the answer. BioTwin brings these computational routes into one research environment, so the next useful investigation can start at the level of detail it needs.
Can the proposed route work within the geometric model?
What does prepared work imply for the next compatible question?
Resolve what the current model cannot settle.
- Test a constraint directly · Use explicit geometry and uncertainty to examine fit and access before committing to an assumed binding route. Molecular fit and access
- Ask another question of a prepared model · Supported landscape queries operate on supplied states and transitions. Reusing that model can avoid regenerating its underlying trajectory for each compatible question. Kinetic landscape queries
- Escalate when detail matters · Use additional sampling, detailed simulation or experiment when the current representation cannot settle the question. Model preparation and validation remain part of the work.
The value is what carries into the next calculation.
In the Lp(a) program, BioTwin compared alternative molecular designs and their supporting assumptions, helping define a candidate program and a focused laboratory verification plan. A revised structural assumption can affect a candidate comparison and the next experiment. BioTwin can query declared relationships to identify which recorded conclusions need review. The query identifies dependencies; it does not automatically rerun or validate the science.
- Follow a changed assumption · See how a chemical input connects to downstream decisions, and where the recorded dependency map stops. Explore the connected example
From a pocket to a plausible mechanism
In the Lp(a) program, BioTwin compared alternative molecular designs and their supporting assumptions, helping define a candidate program and a focused laboratory verification plan.
- Where could we act? · Examine structures, conformations, pockets, interfaces and biological context.
- How could it work? · Investigate chemistry, state changes, recruitment, production or clearance.
- What would change the answer? · Follow the constraints that matter: selectivity, exposure, competing reactions and uncertain inputs.
- What should we test? · Choose an observation that could distinguish the leading explanations.
Compare the reasons, as well as the scores
Native order lattices in 8DB can compare alternatives across declared criteria. When one option has stronger access evidence and another has stronger chemistry evidence, the comparison can retain both for further investigation. This comparison capability and the planning example above have separate demonstrations.
- From a score to a reasoned comparison · See what supports an option, where options remain incomparable and which new evidence could change the choice. Explore the 8DB foundation
Where the difference should show up
LiveDesign already brings models, experimental data and collaboration together. AiiDA already records computational provenance. The useful comparison is which scientific operations are available, how their assumptions connect and what it takes to reach the next decision.
| Research need | Established approach | BioTwin focus | What to evaluate |
|---|---|---|---|
| Bring methods and results together | LiveDesign integrates models and project data for scientific collaboration. | Connect specific operations such as conformational access, reaction geometry and competing kinetics around a candidate mechanism. | Does the combined analysis change which design or experiment you choose? |
| Understand how a result was produced | AiiDA represents data and calculations in a provenance graph and supports reproducible workflows. | Use native scientific representations and declared evidence relationships alongside the result. | Can you inspect the scientific assumptions supporting the candidate comparison and next experiment? |
| Choose between options with different strengths | Teams can compare multiple criteria within existing tools or custom analysis. | 8DB’s native order-lattice operations retain incomparable options under declared criteria; an Lp(a) evidence-frontier demonstration exercises this capability. | Can the comparison identify the missing evidence without concealing a useful alternative in an aggregate score? |
| Choose the computing route | GROMACS supports CPU-only runs; larger or longer simulations can benefit from GPUs and clusters. Reduced-model libraries also answer selected questions efficiently. | Ordinary desktop/laptop deployment with geometric, ensemble and state-model operations alongside connected program evidence. | Compare the same accepted scientific question, including preparation and any detailed follow-up. Local execution alone is not exclusive. |
| Revise a scientific assumption | Provenance systems such as AiiDA retain data/process dependencies for reproducibility and analysis. | Declared relationships connect a structural input, candidate comparison and next experiment. | Inspect what each dependency means scientifically. A queryable dependency is not a demonstration of automatic recalculation. |
These are differences in focus and demonstrated operations, not a claim that other platforms cannot be extended. BioTwin’s planning and order-lattice demonstrations are separate; their deeper integration is in development.
Choose an entry point
Explore a new target, investigate a mechanism, challenge a candidate design or compare intervention strategies. Method selection and the scope of an evaluation depend on the target, available inputs and decision you need to make.
- Scientific capabilities · Find a named operation and the research question it addresses. Scientific capabilities
- Solution modalities · Explore ways to make the desired biological change. Solution Modalities
- Worked investigations · Follow computational work through a result and the questions it leaves open. Evidence
Test BioTwin on a decision your team faces
Use the same target, inputs and research question as your current workflow. Compare the hypotheses worth testing, the evidence supporting them and the work needed to make the decision. That evaluation should determine where BioTwin adds value.