MOTIVATION LETTER The WIPO Fellowship Programme exists to build intellectual property capacity where it is weakest, and that is precisely the gap my venture must cross. I am Eniola Olutogun, a pharmacist turned machine learning engineer, and the sole founder of a pre-seed computational biology venture that predicts drug resistance mutations from protein sequence alone. Our model uses ESM-2 protein language model delta-embeddings combined with ECFP4 drug fingerprints and a Random Forest classifier. On the Platinum benchmark, it achieves an AUROC of 0.804 with a standard deviation of 0.025 under protein-grouped cross-validation, covering 100 percent of mutations tested. Structure-based tools like mCSM-lig cover roughly 18 percent of the same mutation space and achieve an AUROC near 0.70. We exceed that published state of the art while requiring no crystal structure. The venture is not yet incorporated. Proof of concept is validated, but the path to a commercial product runs directly through intellectual property. A protein language model that predicts resistance mutations has patentable claims in model architecture, in the delta-embedding method, and in the specific application to drug resistance screening. Freedom to operate is unclear because the landscape of protein language model patents is crowded and evolving. Licensing terms for foundation model weights, particularly ESM-2, need careful review. These are legal tasks that determine the difference between a publishable method and a defensible product. The WIPO Fellowship offers the structured exposure to patent drafting, prior art analysis, and technology transfer that I cannot get from a startup accelerator. My background is interdisciplinary, a pharmacy degree followed by machine learning engineering, which matches the fellowship's interest in candidates who connect technical depth with IP practice. I am applying from Nigeria, a geography WIPO explicitly seeks to include in global IP capacity building. The fellowship would let me bring that expertise back to a region where biotech founders routinely lose value because they file too late, file in the wrong jurisdiction, or never file at all. My immediate need is the competence to file correctly, license fairly, and navigate the IP landscape of AI in drug discovery. The stipend supports the time to do that work. The training supports the venture. The network supports the next decade. I ask for the fellowship because the venture's bottleneck is not model performance, it is IP strategy, and WIPO is the institution that trains people to solve exactly that problem. RESEARCH STATEMENT The venture's core technical contribution is a method for predicting drug resistance mutations from protein sequence alone, without requiring a crystal structure. The method works as follows. We take an ESM-2 protein language model and compute delta-embeddings, which capture the change in the model's internal representation when a mutation is introduced into the sequence. We concatenate those delta-embeddings with ECFP4 drug fingerprints, which encode the chemical structure of the drug in question. A Random Forest classifier then learns to distinguish resistance-conferring mutations from neutral ones for that specific drug. Validation results are concrete. On the Platinum benchmark, which contains 553 mutations, our method achieves an AUROC of 0.804 with a standard deviation of 0.025 under protein-grouped cross-validation. This grouping is critical because it prevents the model from cheating by memorizing proteins it has already seen. On SKEMPI 2.0, a binding affinity benchmark, the AUROC drops to 0.634, which tells us the method generalizes across tasks but is strongest on resistance prediction specifically. The published state of the art, mCSM-lig, achieves an AUROC near 0.70 on Platinum but covers only about 18 percent of mutations because it requires a crystal structure. Our method covers 100 percent of mutations in the benchmark. That coverage difference is the commercial wedge. Most clinically observed mutations occur in proteins without solved structures. The roadmap has three stages. First, fine-tune ESM-2 on the SKEMPI 3K mutation dataset to improve binding affinity prediction, targeting an AUROC of at least 0.70 on that benchmark. Second, run a pilot with Servier in Suresnes, France, where the tool will be tested on their internal drug resistance questions. Third, convert the pilot into recurring revenue through a software-as-a-service license for pharmaceutical R&D teams. The IP questions are immediate, not downstream. The delta-embedding method is novel and potentially patentable. The combination of protein language model embeddings with drug fingerprints for resistance prediction is a specific claim class. The training pipeline, including the protein-grouped cross-validation scheme, may have trade secret value. The venture must also navigate freedom to operate around existing protein language model patents, and must understand the license terms for ESM-2 weights from Meta. A fellowship at WIPO would give me the training to draft claims, conduct prior art searches, and structure licensing agreements for this specific technology stack. The venture is pre-seed and not yet incorporated. Proof of concept is validated. Revenue is zero. The next milestone is the Servier pilot, and the pilot depends on both technical completion and IP clarity. The fellowship addresses the second dependency directly. ESSAY: DIVERSITY AND GLOBAL IP CAPACITY WIPO's selection criteria emphasize diversity of background and motivation to contribute to economic growth in the home country or region. I am a Nigerian pharmacist who retrained as a machine learning engineer. That path is unusual, and it positions me to build IP capacity in a region where biotech innovation is accelerating but IP literacy is not. Nigeria has a growing computational biology community, but almost no founders who understand patent strategy for AI models. The consequences are visible. Startups publish methods, then watch foreign companies patent the same ideas. Startups sign licensing agreements without understanding what they have given away. Startups fail to file in the right jurisdictions because they do not know the Patent Cooperation Treaty process. The WIPO Fellowship trains exactly this skill set, and I would bring it back not only to my own venture but to the broader ecosystem through mentorship and writing. My venture is a concrete example of why this matters. The method predicts drug resistance mutations from protein sequence alone, covering 100 percent of mutations in the Platinum benchmark versus 18 percent for structure-based tools. That is a patentable method with global relevance for antimicrobial resistance and oncology. If I do not file correctly, the value leaks. If I do, it becomes an asset that can support licensing deals with pharmaceutical companies like Servier and Sanofi. The fellowship is the difference between those two outcomes. ESSAY: MOTIVATION TO CONTRIBUTE TO ECONOMIC GROWTH The venture's commercial model is a software-as-a-service platform for pharmaceutical R&D teams. The target customer is a drug discovery group that needs to know, before clinical trials, which resistance mutations are likely to emerge for a candidate drug. The current alternative is experimental mutagenesis, which takes months and costs hundreds of thousands of dollars. Our method produces a ranked list of likely resistance mutations in hours from sequence data alone. The economic impact has two layers. The first layer is direct revenue from pharmaceutical licenses, starting with the Servier pilot and expanding to other companies. The second layer is the reduction in drug development cost and time, which accelerates the arrival of new antibiotics and antivirals. Antimicrobial resistance is projected to cause 10 million deaths per year by 2050. A tool that helps pharmaceutical companies design drugs that evade resistance earlier in the pipeline has measurable public health value. For Nigeria specifically, the venture creates a template for AI-native biotech companies. It demonstrates that a founder can build a globally competitive computational biology tool from Lagos or Ibadan, without a wet lab, without a university affiliation in Europe or North America, and without a large founding team. The WIPO Fellowship would allow me to model the correct IP behavior for that template, so the next generation of Nigerian biotech founders does not repeat the mistakes of the first generation. CHECKLIST - [ ] Confirm WIPO Fellowship 2026 application deadline on the programme website - [ ] Verify eligibility: first-level university degree (pharmacy degree qualifies) - [ ] Obtain official academic transcripts for pharmacy degree and any ML engineering credentials - [ ] Draft and proofread motivation letter (300-500 words, final version above) - [ ] Draft and proofread research statement (400-600 words, final version above) - [ ] Draft and proofread diversity essay (200-350 words, final version above) - [ ] Draft and proofread economic growth essay (200-350 words, final version above) - [ ] Prepare CV in WIPO-required format, emphasizing pharmacy, ML engineering, and venture milestones - [ ] Obtain two reference letters, one academic and one professional, ideally from Paris-Saclay or Institut Pasteur contacts - [ ] Prepare a one-page summary of venture validation results (AUROC 0.804 on Platinum, 100 percent coverage) - [ ] Prepare a one-page IP landscape summary for protein language models in drug resistance prediction - [ ] Confirm whether WIPO requires a specific application form or online portal submission - [ ] Confirm whether WIPO requires a research proposal or work plan for the fellowship period - [ ] Confirm whether WIPO requires proof of English proficiency - [ ] Submit application before the deadline and retain confirmation receipt EDITOR NOTES - Eligibility risk: WIPO Fellowships typically require a first-level university degree, which the pharmacy degree satisfies, but confirm whether the fellowship requires a postgraduate degree for the specific track; if so, the ML engineering credentials may not qualify without a formal master's. - Fact verification needed: The AUROC of 0.804 on Platinum and 0.634 on SKEMPI 2.0 must be verified against the actual benchmark papers and the venture's own validation runs before submission; WIPO reviewers may ask for methodology details. - Gap: The essays reference mentorship and writing for the Nigerian biotech ecosystem, but no specific prior mentorship or community-building activity is listed in the profile; insert one concrete example of existing ecosystem engagement or be prepared to answer questions about it in the interview. - Gap: The research statement mentions the Servier pilot as the next milestone, but the profile does not specify whether a pilot agreement or letter of intent exists; do not imply a signed agreement unless one exists. - The WIPO Fellowship is not a grant and does not provide seed capital; the application should not imply that fellowship funding will support product development, only that the stipend supports the founder during the IP training period.