Draft — DayOne Accelerator 2026

MOTIVATION LETTER

The Platinum benchmark contains 553 drug-resistance mutations. Structure-based prediction tools can only score about 18 percent of them because they require a resolved three-dimensional protein structure. My sequence-based system, TOPOLOGIX, covers all 553 mutations and reaches an AUROC of 0.804 plus or minus 0.025, outperforming the structure-based baseline mCSM-lig at roughly 0.70. That gap is the venture.

TOPOLOGIX predicts drug-resistance mutations from protein sequence alone. It combines ESM-2 protein language model delta-embeddings with Morgan/ECFP drug fingerprints and a Random Forest classifier. No docking, no crystal structure, no homology model required. A pharmaceutical company can run a resistance screen on any candidate drug the day they have the target sequence, not months later when a structure is solved. This is a software tool for pharma R&D, not a molecule, which matches the DayOne Accelerator focus on techbio infrastructure.

The proof-of-concept is already validated. On the Platinum benchmark, TOPOLOGIX achieves AUROC 0.804 with a standard deviation of 0.025 across cross-validation folds. On SKEMPI 2.0 it reaches 0.634. The system covers 100 percent of mutations in the benchmark, where structure-limited tools cover roughly 18 percent. The comparison is direct and the numbers are published in my preprint.

I came to this problem through a falsified hypothesis. My earlier work tested whether bipartite persistent homology of protein-ligand interface geometry could predict hERG cardiotoxicity. A pre-registered, powered replication showed topological features do not beat a plain descriptor baseline: AUROC 0.8426 versus 0.8782. I then applied the same topological constructs to drug-resistance prediction and found they carried almost no signal, AUROC 0.425 and 0.485 on the Platinum benchmark. Interface geometry is not the driver. That negative result is what pushed me to sequence representations, and TOPOLOGIX is the outcome. I report the failed tests directly because they define the design space.

My background is pharmacology and computation. I hold a B.Pharm from the University of Ibadan with a German equivalent grade of 1.9 and I am enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27. I have built and calibrated ODE models of reward-memory encoding in addiction, Bayesian MCMC parameter estimation pipelines, and a receptor-to-behavior brain-circuit simulation engine with 62 passing tests. I know the drug development pipeline from the pharmacy side and the modeling side.

DayOne Accelerator 2026 targets early-stage healthtech ventures with working prototypes and a clear value proposition for pharma R&D. TOPOLOGIX has a working prototype, a validated benchmark result, and a direct answer to the question of why a pharma company should adopt it: broader coverage and better accuracy than structure-based tools, with a fraction of the computational cost. I am applying to the DayOne network for the mentorship, the pharma ecosystem connections, and the structured path to a first enterprise pilot.

The next milestone is an enterprise pilot with a mid-size pharma or biotech company running TOPOLOGIX on an internal resistance dataset. I am seeking the accelerator's support to make those introductions and to refine the product from research prototype to deployable pipeline.

RESEARCH STATEMENT

TOPOLOGIX is a sequence-based drug-resistance mutation predictor built on protein language model embeddings. The system takes a protein sequence and a drug fingerprint as input and outputs a resistance probability. The architecture is deliberately simple: ESM-2 delta-embeddings capture the mutation's effect on the protein's learned representation, Morgan/ECFP fingerprints encode the drug's chemical structure, and a Random Forest classifier combines the two. Simplicity is a feature. It makes the system fast, interpretable, and easy to integrate into existing pharma pipelines.

The validation is rigorous. On the Platinum benchmark of 553 drug-resistance mutations, TOPOLOGIX achieves AUROC 0.804 with a standard deviation of 0.025. On SKEMPI 2.0, a binding-affinity mutation benchmark, it reaches AUROC 0.634. The structure-based baseline mCSM-lig scores approximately 0.70 on comparable tasks but only covers mutations with resolved structures, roughly 18 percent of the Platinum set. TOPOLOGIX covers 100 percent. The performance gap is not marginal; it is the difference between a tool that works on a fraction of cases and one that works on all of them.

The design was driven by a falsified hypothesis. My prior work tested whether bipartite persistent homology of protein-ligand interface geometry could predict hERG cardiotoxicity. A pre-registered, powered replication found topological features underperform a plain descriptor baseline: AUROC 0.8426 versus 0.8782. I then applied the same topological constructs to drug-resistance prediction. The result was AUROC 0.425 and 0.485 on the Platinum benchmark, essentially no signal. Interface geometry is not the driver of resistance. That negative result redirected the project toward sequence representations, and TOPOLOGIX is the result. The failed experiments are documented in my preprints and inform the current architecture.

The scientific contribution is a settled comparison the literature had never actually run. Published claims about topological descriptors for toxicity and resistance prediction were not backed by direct baseline comparisons. My pre-registered replication and the subsequent resistance study provide those comparisons. The field now knows that interface topology does not beat simple descriptors for these tasks, and that sequence-based representations do.

The product opportunity is concrete. Pharmaceutical companies screen candidate drugs for resistance mutations during development. Structure-based tools require resolved structures, which are often unavailable for novel targets. TOPOLOGIX removes that constraint. A resistance screen can run on sequence data alone, which is available at the earliest stages of target validation. The system is also fast enough for high-throughput screening across large mutation libraries.

The current stage is validated proof-of-concept. The benchmark results are published in my preprint. The next step is an enterprise pilot with a pharma or biotech partner, running TOPOLOGIX on an internal resistance dataset to demonstrate performance on proprietary targets. The technical roadmap includes expanding the training data beyond Platinum and SKEMPI 2.0, adding uncertainty quantification to the predictions, and building a web-based interface for non-expert users.

The founder fit is direct. I am a licensed pharmacist with a B.Pharm from the University of Ibadan and an M.Sc. student in Digital Health at the Hasso Plattner Institute and University of Potsdam. I have built and calibrated computational models across addiction neuroscience, cardiotoxicity prediction, and brain-circuit simulation. I understand the drug development pipeline from the pharmacy floor and the modeling side. TOPOLOGIX is the venture where my pharmacological knowledge and my computational skills converge on a problem pharma companies actually pay to solve.

ESSAY: WHY THIS VENTURE AND WHY NOW

Drug resistance is the reason antibiotics stop working, the reason targeted cancer therapies fail, and the reason antiviral programs get abandoned. Predicting resistance mutations before they emerge in the clinic would let pharmaceutical companies design around them, prioritize analogs, and de-risk development programs. The tools that exist for this task require a resolved three-dimensional protein structure. For most novel targets, that structure does not exist at the point where resistance screening would be most useful.

TOPOLOGIX solves this with sequence alone. The system uses ESM-2 protein language model embeddings to represent the mutation's effect on the protein, Morgan fingerprints to represent the drug, and a Random Forest to combine them. It covers 100 percent of mutations in the Platinum benchmark, where structure-based tools cover roughly 18 percent, and it beats them on accuracy: AUROC 0.804 versus approximately 0.70 for mCSM-lig. The technical risk is retired. The remaining work is productization and pilot validation.

The timing is right for three reasons. First, protein language models like ESM-2 reached maturity in the last three years, making sequence-only resistance prediction feasible for the first time. Second, the pharmaceutical industry is under pressure to reduce R&D costs, and computational tools that replace structure-determination bottlenecks are a direct answer. Third, my own validation work has settled the question of what does not work, interface topology, which means the design space is now clear.

I am applying to DayOne Accelerator 2026 because the program targets early-stage healthtech ventures with working prototypes and a clear value proposition for pharma R&D. TOPOLOGIX has the prototype and the benchmark numbers. What I need from DayOne is the network to reach a first enterprise pilot, the mentorship to turn a research tool into a product, and the structure to execute on a commercial roadmap. The DayOne focus on techbio infrastructure, not molecules, matches TOPOLOGIX exactly.

The alternative to acting now is waiting for a large pharma company or a well-funded academic lab to build the same thing. The benchmark results are public. The architecture is simple. The window is open and I intend to move through it.

CHECKLIST

- [ ] Complete DayOne Accelerator 2026 online application form at the startupticker.ch event page
- [ ] Submit motivation letter (300-500 words, included above)
- [ ] Submit research statement (400-600 words, included above)
- [ ] Submit essay response on venture rationale and timing (200-350 words, included above)
- [ ] Attach TOPOLOGIX preprint with Platinum benchmark results (AUROC 0.804 plus or minus 0.025)
- [ ] Attach preprint documenting the falsified interface-topology hypothesis for hERG cardiotoxicity and drug resistance
- [ ] Prepare one-page executive summary of TOPOLOGIX with benchmark table and architecture diagram
- [ ] Prepare pitch deck (10-12 slides) covering problem, solution, validation, market, and roadmap
- [ ] Verify DayOne Accelerator 2026 funding amount and equity terms before any commitment
- [ ] Confirm whether the program requires a registered company entity or accepts pre-incorporation ventures
- [ ] Confirm whether the program has residency requirements or is fully remote
- [ ] Prepare two references: one academic (Kent Berridge, University of Michigan) and one computational (Samuel Gershman, Harvard)
- [ ] Verify enrollment status at Hasso Plattner Institute for Winter Semester 2026/27 is documented for the application

EDITOR NOTES

- Eligibility risk: DayOne Accelerator may require a registered company entity. Eniola is an independent researcher, not an incorporated founder. Confirm incorporation requirements before submitting; if required, register a Swiss or German entity or confirm a pre-incorporation track exists.
- Funding terms unknown: The program page does not specify funding amount or equity terms. Do not sign anything without legal review. The strategy notes flag this as unspecified, so treat any term sheet as a separate negotiation.
- Verification needed: Confirm the Platinum benchmark AUROC 0.804 plus or minus 0.025 and SKEMPI 2.0 AUROC 0.634 are stated identically in the preprint that will be attached. Also confirm the mCSM-lig baseline of approximately 0.70 is cited correctly from the literature.
- Personal detail gap: The motivation letter references a next milestone of an enterprise pilot. Eniola should insert any existing conversations or expressions of interest from pharma or biotech companies, if any exist, to strengthen the commercial-readiness claim. If no such conversations exist, the letter should say the pilot is the goal of the accelerator engagement, not an ongoing process.
- Honest stage framing: TOPOLOGIX is a validated proof-of-concept, not a product with revenue or customers. The application materials correctly avoid claiming product-market fit or revenue. Keep it that way in any oral presentation or Q&A with DayOne staff.

MOTIVATION LETTER

The Platinum benchmark contains 553 drug-resistance mutations. Structure-based prediction tools can only score about 18 percent of them because they require a resolved three-dimensional protein structure. My sequence-based system, TOPOLOGIX, covers all 553 mutations and reaches an AUROC of 0.804 plus or minus 0.025, outperforming the structure-based baseline mCSM-lig at roughly 0.70. That gap is the venture.

TOPOLOGIX predicts drug-resistance mutations from protein sequence alone. It combines ESM-2 protein language model delta-embeddings with Morgan/ECFP drug fingerprints and a Random Forest classifier. No docking, no crystal structure, no homology model required. A pharmaceutical company can run a resistance screen on any candidate drug the day they have the target sequence, not months later when a structure is solved. This is a software tool for pharma R&D, not a molecule, which matches the DayOne Accelerator focus on techbio infrastructure.

The proof-of-concept is already validated. On the Platinum benchmark, TOPOLOGIX achieves AUROC 0.804 with a standard deviation of 0.025 across cross-validation folds. On SKEMPI 2.0 it reaches 0.634. The system covers 100 percent of mutations in the benchmark, where structure-limited tools cover roughly 18 percent. The comparison is direct and the numbers are published in my preprint.

I came to this problem through a falsified hypothesis. My earlier work tested whether bipartite persistent homology of protein-ligand interface geometry could predict hERG cardiotoxicity. A pre-registered, powered replication showed topological features do not beat a plain descriptor baseline: AUROC 0.8426 versus 0.8782. I then applied the same topological constructs to drug-resistance prediction and found they carried almost no signal, AUROC 0.425 and 0.485 on the Platinum benchmark. Interface geometry is not the driver. That negative result is what pushed me to sequence representations, and TOPOLOGIX is the outcome. I report the failed tests directly because they define the design space.

My background is pharmacology and computation. I hold a B.Pharm from the University of Ibadan with a German equivalent grade of 1.9 and I am enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27. I have built and calibrated ODE models of reward-memory encoding in addiction, Bayesian MCMC parameter estimation pipelines, and a receptor-to-behavior brain-circuit simulation engine with 62 passing tests. I know the drug development pipeline from the pharmacy side and the modeling side.

DayOne Accelerator 2026 targets early-stage healthtech ventures with working prototypes and a clear value proposition for pharma R&D. TOPOLOGIX has a working prototype, a validated benchmark result, and a direct answer to the question of why a pharma company should adopt it: broader coverage and better accuracy than structure-based tools, with a fraction of the computational cost. I am applying to the DayOne network for the mentorship, the pharma ecosystem connections, and the structured path to a first enterprise pilot.

The next milestone is an enterprise pilot with a mid-size pharma or biotech company running TOPOLOGIX on an internal resistance dataset. I am seeking the accelerator's support to make those introductions and to refine the product from research prototype to deployable pipeline.

RESEARCH STATEMENT

TOPOLOGIX is a sequence-based drug-resistance mutation predictor built on protein language model embeddings. The system takes a protein sequence and a drug fingerprint as input and outputs a resistance probability. The architecture is deliberately simple: ESM-2 delta-embeddings capture the mutation's effect on the protein's learned representation, Morgan/ECFP fingerprints encode the drug's chemical structure, and a Random Forest classifier combines the two. Simplicity is a feature. It makes the system fast, interpretable, and easy to integrate into existing pharma pipelines.

The validation is rigorous. On the Platinum benchmark of 553 drug-resistance mutations, TOPOLOGIX achieves AUROC 0.804 with a standard deviation of 0.025. On SKEMPI 2.0, a binding-affinity mutation benchmark, it reaches AUROC 0.634. The structure-based baseline mCSM-lig scores approximately 0.70 on comparable tasks but only covers mutations with resolved structures, roughly 18 percent of the Platinum set. TOPOLOGIX covers 100 percent. The performance gap is not marginal; it is the difference between a tool that works on a fraction of cases and one that works on all of them.

The design was driven by a falsified hypothesis. My prior work tested whether bipartite persistent homology of protein-ligand interface geometry could predict hERG cardiotoxicity. A pre-registered, powered replication found topological features underperform a plain descriptor baseline: AUROC 0.8426 versus 0.8782. I then applied the same topological constructs to drug-resistance prediction. The result was AUROC 0.425 and 0.485 on the Platinum benchmark, essentially no signal. Interface geometry is not the driver of resistance. That negative result redirected the project toward sequence representations, and TOPOLOGIX is the result. The failed experiments are documented in my preprints and inform the current architecture.

The scientific contribution is a settled comparison the literature had never actually run. Published claims about topological descriptors for toxicity and resistance prediction were not backed by direct baseline comparisons. My pre-registered replication and the subsequent resistance study provide those comparisons. The field now knows that interface topology does not beat simple descriptors for these tasks, and that sequence-based representations do.

The product opportunity is concrete. Pharmaceutical companies screen candidate drugs for resistance mutations during development. Structure-based tools require resolved structures, which are often unavailable for novel targets. TOPOLOGIX removes that constraint. A resistance screen can run on sequence data alone, which is available at the earliest stages of target validation. The system is also fast enough for high-throughput screening across large mutation libraries.

The current stage is validated proof-of-concept. The benchmark results are published in my preprint. The next step is an enterprise pilot with a pharma or biotech partner, running TOPOLOGIX on an internal resistance dataset to demonstrate performance on proprietary targets. The technical roadmap includes expanding the training data beyond Platinum and SKEMPI 2.0, adding uncertainty quantification to the predictions, and building a web-based interface for non-expert users.

The founder fit is direct. I am a licensed pharmacist with a B.Pharm from the University of Ibadan and an M.Sc. student in Digital Health at the Hasso Plattner Institute and University of Potsdam. I have built and calibrated computational models across addiction neuroscience, cardiotoxicity prediction, and brain-circuit simulation. I understand the drug development pipeline from the pharmacy floor and the modeling side. TOPOLOGIX is the venture where my pharmacological knowledge and my computational skills converge on a problem pharma companies actually pay to solve.

ESSAY: WHY THIS VENTURE AND WHY NOW

Drug resistance is the reason antibiotics stop working, the reason targeted cancer therapies fail, and the reason antiviral programs get abandoned. Predicting resistance mutations before they emerge in the clinic would let pharmaceutical companies design around them, prioritize analogs, and de-risk development programs. The tools that exist for this task require a resolved three-dimensional protein structure. For most novel targets, that structure does not exist at the point where resistance screening would be most useful.

TOPOLOGIX solves this with sequence alone. The system uses ESM-2 protein language model embeddings to represent the mutation's effect on the protein, Morgan fingerprints to represent the drug, and a Random Forest to combine them. It covers 100 percent of mutations in the Platinum benchmark, where structure-based tools cover roughly 18 percent, and it beats them on accuracy: AUROC 0.804 versus approximately 0.70 for mCSM-lig. The technical risk is retired. The remaining work is productization and pilot validation.

The timing is right for three reasons. First, protein language models like ESM-2 reached maturity in the last three years, making sequence-only resistance prediction feasible for the first time. Second, the pharmaceutical industry is under pressure to reduce R&D costs, and computational tools that replace structure-determination bottlenecks are a direct answer. Third, my own validation work has settled the question of what does not work, interface topology, which means the design space is now clear.

I am applying to DayOne Accelerator 2026 because the program targets early-stage healthtech ventures with working prototypes and a clear value proposition for pharma R&D. TOPOLOGIX has the prototype and the benchmark numbers. What I need from DayOne is the network to reach a first enterprise pilot, the mentorship to turn a research tool into a product, and the structure to execute on a commercial roadmap. The DayOne focus on techbio infrastructure, not molecules, matches TOPOLOGIX exactly.

The alternative to acting now is waiting for a large pharma company or a well-funded academic lab to build the same thing. The benchmark results are public. The architecture is simple. The window is open and I intend to move through it.

CHECKLIST

- [ ] Complete DayOne Accelerator 2026 online application form at the startupticker.ch event page
- [ ] Submit motivation letter (300-500 words, included above)
- [ ] Submit research statement (400-600 words, included above)
- [ ] Submit essay response on venture rationale and timing (200-350 words, included above)
- [ ] Attach TOPOLOGIX preprint with Platinum benchmark results (AUROC 0.804 plus or minus 0.025)
- [ ] Attach preprint documenting the falsified interface-topology hypothesis for hERG cardiotoxicity and drug resistance
- [ ] Prepare one-page executive summary of TOPOLOGIX with benchmark table and architecture diagram
- [ ] Prepare pitch deck (10-12 slides) covering problem, solution, validation, market, and roadmap
- [ ] Verify DayOne Accelerator 2026 funding amount and equity terms before any commitment
- [ ] Confirm whether the program requires a registered company entity or accepts pre-incorporation ventures
- [ ] Confirm whether the program has residency requirements or is fully remote
- [ ] Prepare two references: one academic (Kent Berridge, University of Michigan) and one computational (Samuel Gershman, Harvard)
- [ ] Verify enrollment status at Hasso Plattner Institute for Winter Semester 2026/27 is documented for the application

EDITOR NOTES

- Eligibility risk: DayOne Accelerator may require a registered company entity. Eniola is an independent researcher, not an incorporated founder. Confirm incorporation requirements before submitting; if required, register a Swiss or German entity or confirm a pre-incorporation track exists.
- Funding terms unknown: The program page does not specify funding amount or equity terms. Do not sign anything without legal review. The strategy notes flag this as unspecified, so treat any term sheet as a separate negotiation.
- Verification needed: Confirm the Platinum benchmark AUROC 0.804 plus or minus 0.025 and SKEMPI 2.0 AUROC 0.634 are stated identically in the preprint that will be attached. Also confirm the mCSM-lig baseline of approximately 0.70 is cited correctly from the literature.
- Personal detail gap: The motivation letter references a next milestone of an enterprise pilot. Eniola should insert any existing conversations or expressions of interest from pharma or biotech companies, if any exist, to strengthen the commercial-readiness claim. If no such conversations exist, the letter should say the pilot is the goal of the accelerator engagement, not an ongoing process.
- Honest stage framing: TOPOLOGIX is a validated proof-of-concept, not a product with revenue or customers. The application materials correctly avoid claiming product-market fit or revenue. Keep it that way in any oral presentation or Q&A with DayOne staff.

AI Draft — DayOne Accelerator 2026