> For the complete documentation index, see [llms.txt](https://ersilia.gitbook.io/ersilia-annual-reports/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://ersilia.gitbook.io/ersilia-annual-reports/annual-report-2025/annual-report-2025.md). # Annual Report 2025 ## Introduction The Ersilia Open Source Initiative Foundation (Ersilia) is a non-profit research organization dedicated to developing Artificial Intelligence (AI) methods for infectious disease research. Ersilia works in collaboration with research institutions in the Global South, particularly in Africa, and with global organizations dedicated to open-source data science and antibiotic drug discovery. The foundation's work is based on ten core principles that describe the founders' vision regarding the development needs of the biomedical research fabric in the Global South, in order to reduce the existing inequalities in the field of health. ### Ersilia's ten principles 1. **Science is a driver of socioeconomic development**. It must be done everywhere, including in the poorest regions. 2. **Modern science is rooted in colonialist practices**. This should be acknowledged and reversed. 3. **Research agendas must be dictated by the needs of the people worldwide**, not by a privileged minority or to satisfy private interests. 4. Scientific knowledge should be immediately **free and accessible to all**. 5. **AI and digital technologies** have become a fundamental part of experimental sciences. 6. **Open source software**, if properly funded and maintained, promotes reproducibility and reusability of scientific outputs. 7. **Community building** is at the root of scientific innovation and ensures participation of the global majority. 8. Training and mentoring are integral to scientific development and the **only path to sustainability.** 9. AI requires vast amounts of data. Also in science, **data providers** should be credited and data extractivism avoided. 10. AI is transforming our world — at the risk of deepening existing inequalities. Science is one aspect of society where **AI can bear undisputable good**. ### Ersilia's lines of action Ersilia works on three main pillars: 1. The development of AI models for open-source, patent-free scientific research 2. The implementation of these AI methods in projects led by African researchers 3. The promotion of AI use through capacity-building courses aimed at African scientists ## Strategic Plan 2025-2027 Ersilia's strategic plan for 2025-2027 details the foundation's key actions and objectives over a three-year horizon. Below we provide a brief summary to facilitate tracking the annual report in relation to the priorities described in the plan: ### Priority 1: develop a reference AI platform for infectious disease research The Ersilia Model Hub is the foundation's central open-source platform. It contains a variety of Artificial Intelligence models for infectious disease drug discovery that we deploy daily in Ersilia's internal research as well as in collaborations with researchers around the world. The main objectives regarding the Ersilia Model Hub are: 1. Increase the number of AI models available in the Ersilia Model Hub 2. Improve the accessibility and documentation of the platform 3. Offer an online version for non-expert users 4. Create a precomputed database in order to save computational resources ### Priority 2: new therapeutic opportunities and new targets for neglected diseases Advances in AI for biomedical research are opening new doors to develop treatment modalities that were not possible before, including protein degradation methods and the analysis of -omics data. At Ersilia we want to benefit from these advances and bring them to the field of global health, with the aim of: 1. Identifying new therapeutic targets for neglected tropical diseases 2. Leading the adoption of protein degradation as an antimicrobial method 3. Exploring the chemical space associated with antimicrobial drug discovery, with a special interest in natural products ### Priority 3: offer long-term training to Global South scientists The need for training in the field of AI and data science has become clear to Ersilia during the development of our mission in the early years. For this reason, at Ersilia we focus on two main activities: 1) establishing small local units at research centers that can act as amplifiers in their regions, and 2) developing training courses for scientists (MSc, PhD, PostDoc). For the coming years, our objective is to: 1. Deliver at least one in-person course in the Global South each year 2. Establish a new training program in an incubator format 3. Offer mentorship programs for minorities in the field of science and technology, especially for software development. ### Priority 4: build a community that amplifies our impact and ensures its sustainability Community is the foundation of Ersilia's success, both internationally, with collaborators in the Global South and Global North, and locally in Catalonia, where the foundation has had its offices since 2024. In order to keep building community, our objectives are: 1. Establish new collaborations with Global South institutions 2. Grow a local network of collaborators in Barcelona 3. Build a diverse and inclusive workspace 4. Form an advisory board 5. Diversify funding sources 6. Expand the Ersilia team both technically and operationally ## Roadmap In relation to the priorities described above, the foundation set the following overall objectives: * Priority 1: * Publish the Ersilia Model Hub * Reach 500 models in the hub * Priority 2: * Develop a project in the field of protein degradation as a therapeutic modality * Priority 3: * Offer in-person courses in two countries * Priority 4: * Establish new collaborations with five research centers ## Summary Progress against the set objectives * Priority 1: delayed * Publication of the Ersilia Model Hub in an open repository (BioRxiv) in May 2026 * Current models in the Hub: 200 * Priority 2: achieved * Protein degradation project in *Mycobacterium tuberculosis* in collaboration with Stellenbosch University (2025-2026) * Protein degradation project in *Klebsiella pneumoniae*, part of the Gr-ADI consortium (Stellenbosch University, Rhodes University, Ersilia) (2026-2028) * Priority 3: achieved * Throughout 2025 we delivered in-person courses in South Africa, Kenya, Argentina and Cameroon * Priority 4: achieved * Throughout 2025 we established collaborations with the Global Antibiotic Research and Development Partnership (GARDP, international), IMPM (Cameroon), Stellenbosch University (South Africa), University of Córdoba (Argentina) and DaltonTx (biotech start-up, United Kingdom). ## Ersilia Model Hub The development of the platform has been one of the Foundation's main activities in 2025. Despite the efforts dedicated, it is the only strategic priority where the objectives set for 2025 were not achieved, due to a series of technical and operational difficulties. First, in order to facilitate the maintenance and deployment of the more than 200 AI models available through the Hub, we redesigned the model packaging method, discarding the method used until 2024 (BentoML) and adopting a new implementation developed by the Ersilia team ([Ersilia Pack](https://github.com/ersilia-os/ersilia-pack)). In parallel, we automated the testing of models both at the moment of incorporation and on a regular basis to ensure the correct functioning of the platform ([Maintenance](https://github.com/ersilia-os/ersilia-maintenance) and [Automation](https://github.com/ersilia-os/ersilia-model-workflows)). Finally, we introduced several improvements to the platform, including a Python API that mimics the Command Line Interface and a model versioning system. Regarding model incorporation, the 500 models planned for 2025 were not achieved for two reasons: on the one hand, we adopted a consolidation strategy in which we incorporate several related models as a single model in order to reduce computational needs, and on the other hand, the development of the automated pipeline to curate public data (funded by the Plan Generación Conocimiento 2023 project) is being improved and therefore we discarded the models based on these data in order to renew them. In addition, through the AI2050 program, in collaboration with the H3D Centre (University of Cape Town, South Africa), we developed an online platform ([https://hub.ersilia.io](https://hub.ersilia.io/)) that allows access to our models and obtaining predictions for free and without the need to install our software — a key element in the strategic plan and necessary to achieve the training objectives (details below). All these advances were achieved throughout 2025 despite changes in staffing. From February to October the Lead Software Developer position remained vacant, with the rest of the team covering the necessary functions to keep the project moving forward. All these improvements slowed the achievement of the 2025 objectives (scientific publication and incorporation of 500 models), but they strengthened the infrastructure to achieve these objectives throughout 2026. ## Infectious disease research projects Ersilia developed several research projects in parallel throughout 2025, specifically including a focus on new therapeutic modalities and the study of billion-scale databases (ultra-large screenings), thus achieving the objectives set in the strategic plan for 2025. In summary, scientific productivity is measured through publications and citations. In 2025, with a team of only three researchers, we published the following articles: Lead authorship (first or corresponding author) * Turon\*, Mulubwa\* et al. AI coupled to pharmacometric modelling to tailor malaria and tuberculosis treatment in Africa. Nature Communications (2025) * Turon et al. Addressing infectious diseases in Africa by accelerating drug discovery through data science. Communications Medicine (2025) Others: * Carli et al. Learning and actioning general principles of cancer cell drug sensitivity. Nature Communications (2025) * Poisot et al. Ten quick tips to build a Model Life Cycle. PLOS Computational Biology (2025) * Comajuncosa-Creus et al. Integration of diverse bioactivity data into the Chemical Checker compound universe. Nature Protocols (2025) Below we detail the advances in Ersilia's most notable projects: #### Identification of new anti-tuberculosis strategies In 2025 we started a new collaboration within the framework of the Grand Challenges Africa Drug Discovery Accelerator (GC-ADDA) program, funded by the Bill & Melinda Gates Foundation. Ersilia's role in the project led by Prof. Erick Strauss (Stellenbosch University, South Africa) is the identification of pockets in a family of proteins of the bacterium *Mycobacterium tuberculosis* essential for its survival, the tRNA synthetases. Once these pockets or binding sites are computationally characterized, we can identify molecules that bind to them and, through the BacPROTACs system, recruit the bacterium's own cellular degradation machinery to destroy these proteins, causing its death. This strategy has been developed primarily in the field of cancer and has great potential for infectious diseases. The project continues throughout 2026. #### Automated generation of antimicrobial models One of the major gaps for the adoption of AI methods in the field of infectious diseases is the small amount of existing data. During 2025, we completed the development of a [tool](https://github.com/ersilia-os/chembl-antimicrobial-tasks) to automatically curate ChEMBL, the public database with the most open bioactivity data. With this tool, data ready to train AI models can be quickly obtained (typically, lists of molecules and their associated activity as 0 (inactive) or 1 (active)). We are currently training models based on this tool that will be deployed through the Ersilia Model Hub and used in various research projects. Our goal is to cover more than 15 pathogens identified as risk pathogens for global health by the World Health Organization. #### Ultra-large screening for drugs against highly resistant pathogens In parallel with the development of generative AI methods, which are technically complex and require a lot of computational power, we have focused on the field of ultra-large screening — that is, searching in silico libraries of billions of compounds for molecules active against pathogens of interest. This task is also of high technical complexity and we can perform it efficiently thanks to the Ersilia Model Hub. In order to demonstrate the feasibility of this method, we established a collaboration with the Global Antibiotic Research and Development Partnership (GARDP), a global organization dedicated to advancing antimicrobial drug discovery, and H3D, the main drug discovery center in Africa. Ersilia, using the data provided by GARDP, developed new AI models to search the chemical space and proposed a list of 100 molecules that H3D is testing in its laboratories. Throughout 2026, experimental validations of this research will be obtained, which so far has already resulted in the incorporation of more than 10 new AI models into the Ersilia Model Hub. ## Training 2025 was an important year for the parallel development of several training programs in the Global South: * AI2050 Program: in collaboration with the H3D Foundation (South Africa) we offered a fully funded course for East African researchers in Nairobi, Kenya, following the project started in Ghana in 2024 (30 selected participants). Additionally, in order to offer the training to more researchers, we developed an online course (a reduced version of the in-person course) in which more than 200 scientists registered. The online course is maintained through our YouTube channel for anyone interested. Finally, within the AI2050 program we also delivered specific training for H3D researchers in collaboration with Dr. Collins (MIT). The key element for the success of these trainings has been the development of the online platform of the Ersilia Model Hub. * Merck Schistosomiasis Research Grant: in collaboration with IMPM (Yaoundé, Cameroon) and Northumbria University (UK), during 2025 we collaborated on a three-stage training (biostatistics, bioinformatics and AI) dedicated to IMPM researchers (25 participants). The training project was led by Dr. Justin Komguep and funded by Merck, always in the context of schistosomiasis, a neglected tropical disease. * First LATAM workshop: for the first time, Ersilia offered a training course on AI for drug discovery in Buenos Aires, Argentina, in collaboration with the University of Córdoba and the University of Buenos Aires. This course aims to begin establishing new collaborations on the Latin American continent. More than fifty scientists participated in the training. In summary, Ersilia met the AI capacity-building objectives for Global South scientists set for 2025. The area of action that we paused, unlike in other years, is the online mentorship programs for software engineers. The lack of funding for the Outreachy program and changes in the team's technical staff have hindered the implementation of this program, which we hope to resume throughout 2026. ## Community building Networking — that is, establishing new scientific collaborations as well as with other foundations dedicated to scientist capacity building and technology development — is key to achieving our objectives. At Ersilia we firmly believe that we can only advance research against these diseases if we work together and in the open. In this spirit, during 2025 we established collaborations, as mentioned in the previous paragraphs, with GARDP (international), IMPM (Cameroon), Stellenbosch University (South Africa), University of Córdoba (Argentina), and University of Buenos Aires (Argentina). In addition, we also established a new collaboration with the group of Dr. Patrick Aloy (IRB Barcelona, Spain) and with DaltonTx (biotech start-up, United Kingdom), and continued the existing collaborations with the H3D Centre (South Africa), University of Buea (Cameroon), Wistar Institute (United States) and CeMM (Austria). It is worth highlighting the active participation of Ersilia members in various international networks, including the Climate-Sensitive Infectious Disease Network, Open Life Sciences, Software Sustainability Institute and Fast Forward. Ersilia members also gave talks at international conferences (including LED3 and LUCID Symposium (Belgium), RICIFA (Argentina), IRBB Biomed Conference (Spain), MMV DD4GH workshop (Switzerland)), and it is worth noting the award of a Ramón y Cajal position to Ersilia's Scientific Director, Dr. Miquel Duran-Frigola. ### Funding sources Ersilia's work has been made possible thanks to the following grants, donations and subsidies: Research grants: * Ministerio de Ciencia, Innovación y Tecnología, PID2023-148309OA-I00 funded by MICIU/AEI/10.13039/501100011033 (Agencia Estatal de Innovación grant) * AI2050 Fellowship (Prof. Kelly Chibale, sub-award to Ersilia; Schmidt Sciences) * Mozilla Builders Accelerator (Mozilla) * GC-ADDA Program (subcontract to support the project of Prof. Erick Strauss) Subsidies: * Barcelona Activa 2024 Donations: * Splunk Pledge (Splunk) * HPE Accelerating Impact (Hewlett-Packard Enterprise) * BlackRock Employee-Engagement campaign The 2025 financial report can be consulted at this [link](https://drive.google.com/file/d/18Kx3Zlkrs05mFFBAzDOPf8JDqz4RWPGe/view?usp=sharing). ## Acknowledgements None of this would be possible without the community that supports our mission, whether through volunteering, donations or accompaniment. Ersilia's trustees and founders wholeheartedly thank being surrounded by so many people committed to improving health conditions in the most disadvantaged countries. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://ersilia.gitbook.io/ersilia-annual-reports/annual-report-2025/annual-report-2025.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.