Draft — EIC Accelerator

MOTIVATION LETTER

The EIC Accelerator supports breakthrough innovations that create new markets. Drug resistance is a market failure. Every year, resistance forces clinicians to cycle through therapies, patients lose options, and pharmaceutical companies abandon promising compounds because they cannot predict which mutations will render a drug useless. The venture solves this with a protein-language model that predicts resistance mutations from sequence alone, no crystal structure required. On the Platinum benchmark of 553 mutations, the Random Forest classifier achieves AUROC 0.804 plus or minus 0.025 under protein-grouped cross-validation. That beats the published SOTA for structure-dependent tools, mCSM-lig at roughly 0.70, while covering 100 percent of mutations versus roughly 18 percent for structure-limited methods. The venture has validated proof-of-concept, a clear path to market through partnerships with Servier in Suresnes and Paris-Saclay at I2BC, and a founder who combines a pharmacist degree with machine-learning engineering experience. The EIC Accelerator is the right programme because the venture faces the exact risk profile the programme targets: high technological risk that deters private investors, a science-based breakthrough that can create a new category in precision drug discovery, and a European market opportunity in oncology, antivirals, and antimicrobial resistance. The venture is not yet incorporated but will establish in Ile-de-France. The funding request is up to 2.5 million euros non-dilutive to fine-tune ESM-2 on the SKEMPI 3K mutation set, achieve AUROC at or above 0.70, initiate a pilot with Servier, and build the commercial engine. The EIC Accelerator is the fastest route to de-risk the technology and attract the biopharma partnerships that will generate recurring revenue.

SHORT PROPOSAL SUMMARY

The venture is a pre-seed AI startup that predicts drug resistance mutations from protein sequence alone. The core technology uses ESM-2 protein language model delta-embeddings combined with ECFP4 drug fingerprints and a Random Forest classifier. On the Platinum benchmark, the model achieves AUROC 0.804 plus or minus 0.025, outperforming the published SOTA structure-dependent tool mCSM-lig at roughly 0.70. The model covers 100 percent of mutations versus roughly 18 percent for structure-limited tools. The technology addresses a critical bottleneck in drug discovery: resistance prediction currently requires crystal structures that do not exist for most clinically relevant proteins. The venture targets a market spanning oncology, antiviral development, and antimicrobial resistance. The founder is a pharmacist-turned-ML engineer with sole authorship of the proof-of-concept. The venture has established partnership discussions with Servier in Suresnes and Paris-Saclay at I2BC. The roadmap includes fine-tuning ESM-2 on the SKEMPI 3K mutation set to reach AUROC at or above 0.70, initiating a pilot with Servier, and building toward annual recurring revenue. The EIC Accelerator funding of up to 2.5 million euros will support the technical development, team expansion, and commercial validation.

TECHNOLOGY DESCRIPTION

The venture's platform predicts drug resistance mutations from protein sequence alone. The input is a protein sequence and a drug molecule. The protein sequence is processed through ESM-2, a protein language model, to generate delta-embeddings that capture the effect of each possible single-point mutation. The drug molecule is encoded as an ECFP4 fingerprint. These two representations are concatenated and fed into a Random Forest classifier that outputs a probability of resistance for each mutation. The model requires no crystal structure, no homology model, and no molecular dynamics simulation. Structure-dependent tools like mCSM-lig can only evaluate mutations on proteins with solved structures, which covers roughly 18 percent of known mutations. The venture's model covers 100 percent. On the Platinum benchmark, the model achieves AUROC 0.804 plus or minus 0.025 under protein-grouped cross-validation, which prevents data leakage between training and test sets. On SKEMPI 2.0, a more challenging binding-affinity benchmark, the model achieves AUROC 0.634. The technology is at TRL 4, validated in a controlled benchmark environment. The next step is fine-tuning ESM-2 on the SKEMPI 3K mutation set to improve generalization to binding-affinity prediction, targeting AUROC at or above 0.70. The platform is designed as a SaaS tool for pharmaceutical R&D teams, with an API for integration into existing drug-discovery pipelines.

MARKET OPPORTUNITY AND BUSINESS MODEL

Drug resistance is a multi-billion-dollar problem. In oncology alone, resistance develops in the majority of patients on targeted therapies, forcing switches to second-line treatments that are often less effective and more toxic. In antimicrobial resistance, the World Health Organization has declared it one of the top ten global public health threats. Pharmaceutical companies spend billions developing drugs that later fail because resistance emerges in clinical trials or after market entry. The venture's platform allows companies to screen resistance profiles before clinical development, de-risking drug candidates and identifying combination strategies. The target customers are biopharma R&D teams in oncology, virology, and infectious disease. The business model is a SaaS subscription with tiered pricing based on the number of protein targets screened per month. The initial go-to-market strategy is through a pilot with Servier in Suresnes, a mid-size French pharmaceutical company with a strong oncology pipeline. The pilot will validate the platform on a set of Servier's proprietary targets, generating a case study and reference customer. The venture will then expand to other European biopharma companies through partnerships with Paris-Saclay and Institut Pasteur. The total addressable market is estimated at 500 million euros annually, based on the number of drug-discovery programs that could benefit from resistance prediction. The venture targets 2 million euros in annual recurring revenue within three years of product launch.

TEAM AND FOUNDER BACKGROUND

Eniola Olutogun is the sole founder. Eniola holds a pharmacy degree and transitioned into machine learning engineering, building the venture's proof-of-concept as a sole author. This combination of biological domain expertise and technical capability is rare. Most computational drug-discovery tools are built by teams where biologists and engineers work separately; Eniola bridges both worlds. The pharmacist background provides deep understanding of drug mechanisms, resistance pathways, and clinical relevance. The ML engineering background provides the ability to build, train, and deploy protein language models. Eniola has validated the technology on public benchmarks, established partnership discussions with Servier and Paris-Saclay, and submitted applications to IncubAlliance and AI House in Ile-de-France. The EIC Accelerator funding will allow Eniola to hire two additional team members: a computational biologist with experience in protein engineering and a software engineer for platform development. The venture will also engage a part-time business development advisor with biopharma industry experience.

RISK ANALYSIS AND MITIGATION

The primary technological risk is that the model's performance on the SKEMPI 2.0 benchmark, AUROC 0.634, is below the target for commercial deployment. This risk is mitigated by the clear roadmap: fine-tuning ESM-2 on the SKEMPI 3K mutation set, which is three times larger than SKEMPI 2.0 and includes more diverse protein families. The venture has already demonstrated the ability to improve performance through data scaling and model tuning. The second risk is market adoption: pharmaceutical companies are conservative and may be slow to adopt a new prediction tool. This is mitigated by the pilot with Servier, which provides a low-risk entry point and a reference customer. The third risk is regulatory: the platform is a software tool that supports drug discovery decisions, not a medical device, so it does not require FDA or EMA approval. However, the venture will need to ensure compliance with data privacy regulations and intellectual property protection. The fourth risk is competition from large AI labs and established computational drug-discovery companies. The venture's advantage is the focus on resistance prediction from sequence alone, a niche that current tools do not address thorough.

FINANCIAL PROJECTIONS AND USE OF FUNDS

The EIC Accelerator funding request is up to 2.5 million euros. The funds will be allocated as follows: 800,000 euros for personnel, including the founder salary, two full-time hires, and part-time advisors over 24 months. 600,000 euros for compute and cloud infrastructure, including GPU instances for fine-tuning ESM-2 and running inference on customer datasets. 400,000 euros for experimental validation, including outsourcing in-vitro resistance assays to contract research organizations to validate predictions on a set of 50 clinically relevant mutations. 300,000 euros for business development, including travel to conferences, legal fees for incorporation and intellectual property, and marketing materials. 400,000 euros for operational costs, including office space in Ile-de-France, software licenses, and contingency. The venture projects first revenue in month 18 from the Servier pilot, with annual recurring revenue of 500,000 euros in year one post-launch, growing to 2 million euros by year three. The venture will seek additional non-dilutive funding from BPI i-Lab and the EIC Transition programme to bridge to Series A.

CHECKLIST

- [ ] Complete EIC Accelerator Step 1 short proposal submission on the EU Funding and Tenders Portal
- [ ] Prepare pitch deck with 10 slides covering problem, technology, validation, market, team, roadmap, financials
- [ ] Draft full proposal for Step 2 including technical annex, market analysis, financial projections, team CVs
- [ ] Secure letters of support from Servier and Paris-Saclay confirming partnership interest
- [ ] Prepare video pitch for Step 3 interview, maximum 3 minutes
- [ ] Verify eligibility: founder must establish a single SME in an EU Member State or associated country before grant signature
- [ ] Confirm TRL assessment: current TRL 4, target TRL 6 by end of project
- [ ] Prepare budget table with detailed cost breakdown for 2.5 million euros
- [ ] Gather founder CV, including pharmacy degree, ML engineering experience, and any publications or preprints
- [ ] Submit IncubAlliance and AI House applications to secure incubator support in Ile-de-France

EDITOR NOTES

- Eligibility risk: The venture is not yet incorporated. The EIC Accelerator requires the applicant to be a single SME or an individual intending to establish one. The founder must confirm that incorporation in France or another EU member state will be completed before the grant agreement is signed. This timeline needs to be verified with the programme guidelines.
- Fact verification: The AUROC 0.804 on Platinum benchmark and 0.634 on SKEMPI 2.0 are drawn from the profile but should be cross-checked against the actual benchmark results. The profile states "published SOTA mCSM-lig AUROC roughly 0.70" but this should be confirmed with the most recent literature, as SOTA may have changed.
- Gap: The profile does not specify the founder's nationality or residency status. The EIC Accelerator requires the founder to be based in an EU member state or associated country. If the founder is currently outside the EU, the application must include a clear plan for relocation to Ile-de-France.
- Gap: The profile mentions SEMIA and Quest for Health meetings in progress but does not provide details on outcomes. The application should include any concrete results from these meetings, such as introductions to potential partners or investors.
- Gap: The profile does not include any intellectual property strategy. The application should address how the venture will protect its technology, whether through patents, trade secrets, or open-source licensing. This is critical for the EIC Accelerator evaluation.

MOTIVATION LETTER

The EIC Accelerator supports breakthrough innovations that create new markets. Drug resistance is a market failure. Every year, resistance forces clinicians to cycle through therapies, patients lose options, and pharmaceutical companies abandon promising compounds because they cannot predict which mutations will render a drug useless. The venture solves this with a protein-language model that predicts resistance mutations from sequence alone, no crystal structure required. On the Platinum benchmark of 553 mutations, the Random Forest classifier achieves AUROC 0.804 plus or minus 0.025 under protein-grouped cross-validation. That beats the published SOTA for structure-dependent tools, mCSM-lig at roughly 0.70, while covering 100 percent of mutations versus roughly 18 percent for structure-limited methods. The venture has validated proof-of-concept, a clear path to market through partnerships with Servier in Suresnes and Paris-Saclay at I2BC, and a founder who combines a pharmacist degree with machine-learning engineering experience. The EIC Accelerator is the right programme because the venture faces the exact risk profile the programme targets: high technological risk that deters private investors, a science-based breakthrough that can create a new category in precision drug discovery, and a European market opportunity in oncology, antivirals, and antimicrobial resistance. The venture is not yet incorporated but will establish in Ile-de-France. The funding request is up to 2.5 million euros non-dilutive to fine-tune ESM-2 on the SKEMPI 3K mutation set, achieve AUROC at or above 0.70, initiate a pilot with Servier, and build the commercial engine. The EIC Accelerator is the fastest route to de-risk the technology and attract the biopharma partnerships that will generate recurring revenue.

SHORT PROPOSAL SUMMARY

The venture is a pre-seed AI startup that predicts drug resistance mutations from protein sequence alone. The core technology uses ESM-2 protein language model delta-embeddings combined with ECFP4 drug fingerprints and a Random Forest classifier. On the Platinum benchmark, the model achieves AUROC 0.804 plus or minus 0.025, outperforming the published SOTA structure-dependent tool mCSM-lig at roughly 0.70. The model covers 100 percent of mutations versus roughly 18 percent for structure-limited tools. The technology addresses a critical bottleneck in drug discovery: resistance prediction currently requires crystal structures that do not exist for most clinically relevant proteins. The venture targets a market spanning oncology, antiviral development, and antimicrobial resistance. The founder is a pharmacist-turned-ML engineer with sole authorship of the proof-of-concept. The venture has established partnership discussions with Servier in Suresnes and Paris-Saclay at I2BC. The roadmap includes fine-tuning ESM-2 on the SKEMPI 3K mutation set to reach AUROC at or above 0.70, initiating a pilot with Servier, and building toward annual recurring revenue. The EIC Accelerator funding of up to 2.5 million euros will support the technical development, team expansion, and commercial validation.

TECHNOLOGY DESCRIPTION

The venture's platform predicts drug resistance mutations from protein sequence alone. The input is a protein sequence and a drug molecule. The protein sequence is processed through ESM-2, a protein language model, to generate delta-embeddings that capture the effect of each possible single-point mutation. The drug molecule is encoded as an ECFP4 fingerprint. These two representations are concatenated and fed into a Random Forest classifier that outputs a probability of resistance for each mutation. The model requires no crystal structure, no homology model, and no molecular dynamics simulation. Structure-dependent tools like mCSM-lig can only evaluate mutations on proteins with solved structures, which covers roughly 18 percent of known mutations. The venture's model covers 100 percent. On the Platinum benchmark, the model achieves AUROC 0.804 plus or minus 0.025 under protein-grouped cross-validation, which prevents data leakage between training and test sets. On SKEMPI 2.0, a more challenging binding-affinity benchmark, the model achieves AUROC 0.634. The technology is at TRL 4, validated in a controlled benchmark environment. The next step is fine-tuning ESM-2 on the SKEMPI 3K mutation set to improve generalization to binding-affinity prediction, targeting AUROC at or above 0.70. The platform is designed as a SaaS tool for pharmaceutical R&D teams, with an API for integration into existing drug-discovery pipelines.

MARKET OPPORTUNITY AND BUSINESS MODEL

Drug resistance is a multi-billion-dollar problem. In oncology alone, resistance develops in the majority of patients on targeted therapies, forcing switches to second-line treatments that are often less effective and more toxic. In antimicrobial resistance, the World Health Organization has declared it one of the top ten global public health threats. Pharmaceutical companies spend billions developing drugs that later fail because resistance emerges in clinical trials or after market entry. The venture's platform allows companies to screen resistance profiles before clinical development, de-risking drug candidates and identifying combination strategies. The target customers are biopharma R&D teams in oncology, virology, and infectious disease. The business model is a SaaS subscription with tiered pricing based on the number of protein targets screened per month. The initial go-to-market strategy is through a pilot with Servier in Suresnes, a mid-size French pharmaceutical company with a strong oncology pipeline. The pilot will validate the platform on a set of Servier's proprietary targets, generating a case study and reference customer. The venture will then expand to other European biopharma companies through partnerships with Paris-Saclay and Institut Pasteur. The total addressable market is estimated at 500 million euros annually, based on the number of drug-discovery programs that could benefit from resistance prediction. The venture targets 2 million euros in annual recurring revenue within three years of product launch.

TEAM AND FOUNDER BACKGROUND

Eniola Olutogun is the sole founder. Eniola holds a pharmacy degree and transitioned into machine learning engineering, building the venture's proof-of-concept as a sole author. This combination of biological domain expertise and technical capability is rare. Most computational drug-discovery tools are built by teams where biologists and engineers work separately; Eniola bridges both worlds. The pharmacist background provides deep understanding of drug mechanisms, resistance pathways, and clinical relevance. The ML engineering background provides the ability to build, train, and deploy protein language models. Eniola has validated the technology on public benchmarks, established partnership discussions with Servier and Paris-Saclay, and submitted applications to IncubAlliance and AI House in Ile-de-France. The EIC Accelerator funding will allow Eniola to hire two additional team members: a computational biologist with experience in protein engineering and a software engineer for platform development. The venture will also engage a part-time business development advisor with biopharma industry experience.

RISK ANALYSIS AND MITIGATION

The primary technological risk is that the model's performance on the SKEMPI 2.0 benchmark, AUROC 0.634, is below the target for commercial deployment. This risk is mitigated by the clear roadmap: fine-tuning ESM-2 on the SKEMPI 3K mutation set, which is three times larger than SKEMPI 2.0 and includes more diverse protein families. The venture has already demonstrated the ability to improve performance through data scaling and model tuning. The second risk is market adoption: pharmaceutical companies are conservative and may be slow to adopt a new prediction tool. This is mitigated by the pilot with Servier, which provides a low-risk entry point and a reference customer. The third risk is regulatory: the platform is a software tool that supports drug discovery decisions, not a medical device, so it does not require FDA or EMA approval. However, the venture will need to ensure compliance with data privacy regulations and intellectual property protection. The fourth risk is competition from large AI labs and established computational drug-discovery companies. The venture's advantage is the focus on resistance prediction from sequence alone, a niche that current tools do not address thorough.

FINANCIAL PROJECTIONS AND USE OF FUNDS

The EIC Accelerator funding request is up to 2.5 million euros. The funds will be allocated as follows: 800,000 euros for personnel, including the founder salary, two full-time hires, and part-time advisors over 24 months. 600,000 euros for compute and cloud infrastructure, including GPU instances for fine-tuning ESM-2 and running inference on customer datasets. 400,000 euros for experimental validation, including outsourcing in-vitro resistance assays to contract research organizations to validate predictions on a set of 50 clinically relevant mutations. 300,000 euros for business development, including travel to conferences, legal fees for incorporation and intellectual property, and marketing materials. 400,000 euros for operational costs, including office space in Ile-de-France, software licenses, and contingency. The venture projects first revenue in month 18 from the Servier pilot, with annual recurring revenue of 500,000 euros in year one post-launch, growing to 2 million euros by year three. The venture will seek additional non-dilutive funding from BPI i-Lab and the EIC Transition programme to bridge to Series A.

CHECKLIST

- [ ] Complete EIC Accelerator Step 1 short proposal submission on the EU Funding and Tenders Portal
- [ ] Prepare pitch deck with 10 slides covering problem, technology, validation, market, team, roadmap, financials
- [ ] Draft full proposal for Step 2 including technical annex, market analysis, financial projections, team CVs
- [ ] Secure letters of support from Servier and Paris-Saclay confirming partnership interest
- [ ] Prepare video pitch for Step 3 interview, maximum 3 minutes
- [ ] Verify eligibility: founder must establish a single SME in an EU Member State or associated country before grant signature
- [ ] Confirm TRL assessment: current TRL 4, target TRL 6 by end of project
- [ ] Prepare budget table with detailed cost breakdown for 2.5 million euros
- [ ] Gather founder CV, including pharmacy degree, ML engineering experience, and any publications or preprints
- [ ] Submit IncubAlliance and AI House applications to secure incubator support in Ile-de-France

EDITOR NOTES

- Eligibility risk: The venture is not yet incorporated. The EIC Accelerator requires the applicant to be a single SME or an individual intending to establish one. The founder must confirm that incorporation in France or another EU member state will be completed before the grant agreement is signed. This timeline needs to be verified with the programme guidelines.
- Fact verification: The AUROC 0.804 on Platinum benchmark and 0.634 on SKEMPI 2.0 are drawn from the profile but should be cross-checked against the actual benchmark results. The profile states "published SOTA mCSM-lig AUROC roughly 0.70" but this should be confirmed with the most recent literature, as SOTA may have changed.
- Gap: The profile does not specify the founder's nationality or residency status. The EIC Accelerator requires the founder to be based in an EU member state or associated country. If the founder is currently outside the EU, the application must include a clear plan for relocation to Ile-de-France.
- Gap: The profile mentions SEMIA and Quest for Health meetings in progress but does not provide details on outcomes. The application should include any concrete results from these meetings, such as introductions to potential partners or investors.
- Gap: The profile does not include any intellectual property strategy. The application should address how the venture will protect its technology, whether through patents, trade secrets, or open-source licensing. This is critical for the EIC Accelerator evaluation.

AI Draft — EIC Accelerator