MOTIVATION LETTER Servier’s Spartners incubator sits at Paris-Saclay, the exact ecosystem where my venture’s named partners, Servier, Paris-Saclay I2BC, and Institut Pasteur, already operate. My venture predicts drug resistance mutations from protein sequence alone, using an ESM-2 protein language model combined with ECFP4 drug fingerprints and a Random Forest classifier. On the Platinum benchmark of 553 mutations, the model achieves an AUROC of 0.804 plus or minus 0.025 under protein-grouped cross-validation, beating the published SOTA of mCSM-lig at roughly 0.70. More critically, the method covers 100 percent of mutations in the test set, while structure-limited tools cover only about 18 percent. This means my platform can predict resistance for targets where no crystal structure exists, the majority of the proteome. Servier’s priority areas are oncology and neurology. Both fields face a common bottleneck: targeted therapies fail when tumors or neurons acquire resistance mutations that are invisible to current computational tools. My venture directly addresses this unmet need. The proof-of-concept is validated. The roadmap is concrete: fine-tune ESM-2 on the SKEMPI 3K mutation set to push AUROC above 0.70, then initiate a pilot with Servier at Suresnes. The Spartners incubator provides the lab access, biopharma network, and collaborative ecosystem to execute that pilot within twelve months. I am a pharmacist turned machine learning engineer, sole founder, and sole author of the venture. I have submitted applications to IncubAlliance and AI House, and a meeting with SEMIA and Quest for Health is in progress. The WILCO One BioTech cohort begins October 2026. Spartners is the missing piece: a Servier-anchored incubator at Paris-Saclay that can accelerate the transition from benchmark validation to a commercial pilot with a named pharma partner. The venture is pre-seed, not yet incorporated, and seeking non-dilutive grants between 30,000 and 2.5 million euros. Spartners fits this stage exactly. The programme’s evaluation criteria, scientific innovation, feasibility, team expertise, alignment with Servier’s strategic areas, match my venture’s strengths. I am applying for the Startup track. RESEARCH STATEMENT Drug resistance is the single largest cause of targeted therapy failure in oncology and neurology. Current computational methods for predicting resistance mutations depend on high-resolution protein crystal structures, which exist for fewer than 20 percent of clinically relevant targets. This structural bottleneck leaves the majority of the proteome uncharacterized for resistance risk. My venture eliminates that bottleneck. The core technology uses ESM-2 protein language model delta-embeddings, a representation of how a mutation changes the protein’s internal language, combined with ECFP4 drug fingerprints and a Random Forest classifier. No crystal structure is required. The model was validated on the Platinum benchmark, a standard set of 553 mutations across diverse protein targets. Under protein-grouped cross-validation, which prevents data leakage between similar mutations, the model achieved an AUROC of 0.804 plus or minus 0.025. This exceeds the published SOTA for structure-dependent tools like mCSM-lig, which reports an AUROC of approximately 0.70. On the SKEMPI 2.0 benchmark, which includes binding affinity changes from mutations, the model achieved an AUROC of 0.634, indicating room for improvement on protein-protein interaction data. The key advantage is coverage. Structure-dependent tools can only score mutations on proteins with solved crystal structures, covering roughly 18 percent of known mutations. My method scores 100 percent of mutations, because protein language model embeddings are computed from sequence alone. This opens resistance prediction for thousands of targets in oncology, virology, and antimicrobial resistance that have never been structurally characterized. The immediate technical goal is to fine-tune ESM-2 on the SKEMPI 3K mutation set, which contains approximately 3,000 mutations from protein-protein interaction data. This should push the AUROC on SKEMPI 2.0 above 0.70, matching the current SOTA for structure-dependent tools on that benchmark. Once achieved, the platform will be ready for a pilot with Servier, testing resistance predictions against their experimental validation pipeline at Suresnes. The long-term research direction is to extend the model to predict not just binary resistance versus sensitivity, but the degree of resistance and the most effective alternative drug for each resistant variant. This would turn the platform from a diagnostic tool into a treatment recommendation engine. SHORT ESSAY: WHY SPARTNERS Spartners is the only incubator that combines three elements my venture needs: a direct connection to Servier’s R&D, a physical location at Paris-Saclay where my named partners are based, and a biopharma network that can accelerate a pre-seed startup toward a commercial pilot. I have already named Servier as a key partner in my venture’s roadmap. The Spartners programme makes that partnership operational by providing lab access, mentorship from Servier scientists, and integration into the BioLabs ecosystem. The timing is precise. My proof-of-concept is validated. The next step is a pilot with Servier. Spartners offers the shortest path from benchmark to pilot. The incubator’s focus on oncology and neurology aligns exactly with my venture’s target indications. The evaluation criteria, scientific innovation, feasibility, team expertise, alignment with Servier’s strategic areas, match my venture’s current stage and strengths. I am a sole founder, which makes network access critical. Spartners provides that network without requiring me to relocate or incorporate before acceptance. The programme’s pre-seed focus means I can join as an unincorporated venture and use the incubator resources to incorporate in France, apply for BPI i-Lab and EIC Accelerator grants, and build the team needed for the Servier pilot. SHORT ESSAY: TEAM AND CAPABILITY I am a pharmacist with a Master’s degree in machine learning from a top European programme. My background combines deep domain knowledge in pharmacology, drug mechanisms, resistance pathways, clinical translation, with hands-on engineering in protein language models, molecular fingerprints, and ensemble classifiers. I built the entire proof-of-concept alone: data preprocessing, model architecture, benchmark evaluation, and statistical validation. The venture is pre-seed and sole-founder. This is a risk, but it is also an advantage at this stage. I can make technical and strategic decisions without coordination overhead. The model’s performance, AUROC 0.804 on Platinum, 100 percent mutation coverage, was achieved by a single engineer working with publicly available data and open-source models. With incubator resources, a team, and access to Servier’s experimental platforms, the same approach will scale to clinical-grade accuracy. I have submitted applications to IncubAlliance and AI House, and a meeting with SEMIA and Quest for Health is scheduled. The WILCO One BioTech programme starts in October 2026. These parallel tracks show that I am actively building the venture’s support infrastructure. Spartners is the most strategically aligned piece of that infrastructure. CHECKLIST - [ ] Motivation letter, 300-500 words, tailored to Spartners - [ ] Research statement, 400-600 words, describing technology and validation - [ ] Short essay: Why Spartners, 200-350 words - [ ] Short essay: Team and capability, 200-350 words - [ ] Proof-of-concept data: AUROC 0.804 on Platinum, 0.634 on SKEMPI 2.0 - [ ] Named partner letters or emails: Servier, Paris-Saclay I2BC, Institut Pasteur, Sanofi - [ ] Meeting confirmation: SEMIA / Quest for Health - [ ] Application receipts: IncubAlliance, AI House - [ ] WILCO One BioTech cohort confirmation or application proof - [ ] Founder CV: Eniola Olutogun, pharmacist, ML engineer - [ ] Venture pitch deck or one-pager - [ ] Incorporation plan for France / Ile-de-France EDITOR NOTES - Eligibility risk: Spartners may require the startup to be incorporated. The profile says not yet incorporated. Verify the programme’s incorporation requirement and, if needed, prepare a French SAS incorporation timeline. - Fact to verify: Confirm that Servier’s Suresnes site is the same location as the Spartners incubator. The profile says Spartners is at Paris-Saclay; Servier’s R&D is in Suresnes. Clarify the physical relationship. - Gap: The profile does not mention any co-founders, advisors, or scientific board. For a sole-founder biotech venture, the application may need a letter of support from a named partner scientist (e.g., at I2BC or Institut Pasteur) to strengthen the team section. - Gap: The profile lists Sanofi as a named partner but does not describe the relationship. If Sanofi is a real contact, include a brief sentence in the team essay. If not, remove Sanofi from the application to avoid overclaiming. - Fact to verify: Confirm that the SKEMPI 3K dataset is publicly available and that fine-tuning ESM-2 on it is technically feasible within the incubator timeline. The roadmap assumes this, but the applicant should confirm data access and compute requirements.