AI Clinic for Global Health Researchers
Funding Area: Capacity building / Construcción de capacidad
Problem Statement: Our goal is to remove the inequalities in Global Health, where the diseases that affect the global majority are systematically deprioritized and underfunded. Six of the top ten causes of death in low income countries (LIC) are due to infections, but less than 15% of the drugs in current development tackle these diseases, effectively neglecting the needs of the 3.3 billion people living in low-resource countries. New, affordable medicines for LIC-endemic diseases, like malaria, are required to ameliorate this situation. Data science and, in particular, artificial intelligence (AI) are transforming the field of drug discovery by reducing the costs and number of necessary experiments, making them a unique asset for low-resource settings where, traditionally, the costs of healthcare research and innovation have been prohibitive. However, adoption of AI tools in LMIE remains low, due to a combination of barriers (lack of training, low number of AI tools tailored to infectious diseases, insufficient computing infrastructure and expensive software licenses). Our goal is to equip those laboratories and universities in LMIE with open source state-of-the-art tools to conduct their scientific research. We aim to achieve this by 1) developing open source, ready-to-use AI methods oriented towards infectious diseases 2) providing capacity building and training to local scientists and 3) prioritizing in-country research and local implementation of our assets.
Proposed Activities: Our goal is to pilot an AI Clinic that offers a series of pro-bono consultations to experimental researchers with little or no knowledge of AI tools. Our aim is to guide and support them in identifying the best open source, AI-based solutions to expedite their research. This project directly responds to the needs we have identified over three years of fieldwork at Ersilia. Non-expert scientists hesitate to incorporate AI-driven decisions into their routine, despite the potential benefits, such as reducing the number of necessary samples for an experiment, preventing failures in advanced project stages, or aiding in the critical choices of experiment planning. In such situations, the need is for mentorship and guidance. Without readily available AI experts at their institutions to provide this type of support many AI tools being developed will remain under or misused. This project is inspired by the Research Software Health Check by the Software Sustainability Institute, of which our CEO is a Fellow, and from which we have benefited at Ersilia to evaluate and improve our open source platform of AI models for infectious diseases (https://ersilia.io/model-hub). The AI Clinic will have two stages: 1) an online questionnaire where interested scientists describe their specific research issues and their goals for implementing AI and 2) an evaluation of the selected research projects consisting of a more comprehensive questionnaire, three 1:1 meetings with Ersilia’s CSO and asynchronous work by the Ersilia team to investigate and provide support to the specific needs of the project. At the end of the consultation, the researcher will obtain an assessment of the implementation of a specific AI method, suggestions on which OS AI tools would be most appropriate in a specific project and guidance in the use of those, including Ersilia-developed AI software. In this pilot project we aim to support 5 consultations from scientists in the Global South, offering the possibility of participating in the program either in English or Spanish. If successful, we will write a case-study with each participant and apply for further funding to run the program three times a year, with rolling calls every four months. In line with Ersilia’s mission, we will prioritize projects focused on infectious diseases. This project stands to make a substantial contribution to global health research by democratizing access to AI knowledge to the scientific community, one of the collectives whose work could be most transformed by AI. This will enhance interdisciplinary collaborations, leading to more rapid development of treatments for infectious diseases, enhancing the credibility of AI methods for scientific research and promoting an open source culture.
Openness: The project is open in several aspects: First, all the AI and data science tools used, suggested or evaluated within the AI Clinic will be open source, in line with Ersilia’s work. Second, regarding the sharing of the project outputs openly, we aim to achieve two milestones: writing a case study with each participant and sharing it via social media (Medium, Twitter and LinkedIn) to inspire others that might be pursuing similar work, and also, to benefit the whole community, if within the AI Clinic there are specific tools identified that could be of use but are not easily accessible, the Ersilia team will incorporate them to their catalog of open tools, the Ersilia Model Hub. In addition, any tool or model specifically developed within the project will also be released openly. Third, the participants of the AI Clinic will be welcome to join the Ersilia Slack channel, which currently has > 100 participants from more than 10 countries, to continue building community and networking with other scientists and software engineers in this space. Moreover, those scientists who present a project but are not selected for the consultation will be also added to the Slack channel, to allow them to get support from the broader Ersilia community as well as other participants in the program. Overall, we aim to leverage the AI Clinic to establish a network of infectious disease researchers, better understand their needs and craft our tools and next projects to provide answers to those.
Challenges: Since it is a pilot project and the first time we run it, the main challenge will be to encourage a sufficient number of applications. We aim to use our existing network as well as other initiatives, like the Open Life Sciences, to reach a broad audience. We will create the advertising material both in Spanish and in English, to lower the barrier to access to spanish-speaking scientists, and encourage participation from that community. Our target audience are PhD students, postdoctoral researchers and principal investigators, and the program will run in both languages. Another challenge will be to ensure the objectives of the AI Clinic are clear, and to adjust the project expectations of the participants (this is intended to be a first implementation support, not a fully crafted scientific project). We will devise the entry questionnaire so that we can focus the objective of the consultancy to a specific project that can be tackled within the allocated time frame. The third main challenge we anticipate is the continuation of the project upon finishing the AI Clinic. This will be mitigated by ensuring the tools used or recommended are implemented in a user-friendly manner (either they already are, or Ersilia scientists will incorporate them in the array of open source tools we currently offer), as well as by building a community around the program, leveraging Ersilia’s existing one, so that researchers have other peers to support their work, discuss and advance together.
Neglectedness: We have crafted this program based on our own 3-year experience in the needs of the community we serve, scientists in LMIE. We are applying for funding for this pilot project to the IOI Infrastructure Fund as well as the European AI Fund, though the latter is European-oriented and most of the work we do should target researchers in that geographical location. The EuropeanAI Fund will not be resolved until end of the year. If both were granted, we would be able to incorporate a European cohort to the project, increasing the networking opportunities for the participants. Obtaining seed funding for piloting this project is challenging because i) funders in science are mostly looking for projects that produce new tools or new research, rather than projects that support the re-use of existing assets and ii) most of the AI development, and therefore, funding for it, is directed to well-known institutions in the Global North, rather than in our target geographical areas. We are confident that if we are able to build solid case-studies from this pilot project, we will then be able to apply to larger funding to organizations such as the Bill and Melinda Gates Foundation, the Wellcome Trust or Europe Horizon to offer the program three times a year, but this might prove difficult without this first pilot.
Success: The success measurement will start from the launch of the program. The first metric we will measure is the number of applications for the pilot AI Clinic, and the diversity of those in terms of geographical location, level of expertise in AI, state of the proposed project and career stage of the applicants. This will be crucial for future rounds of the program to revise our dissemination tools and improve the accessibility and participation in the program. Once the 5 pilot projects are selected, we will measure: i)Number of AI tools / methods implemented as a result of the AI Clinics ii) Number of open source AI tools / methods incorporated within the Ersilia infrastructure to facilitate its use iii) New documentation created around open source tools, based on the needs of the participants iv) Level of satisfaction with the service from the participants (measured by a beginning and end of program survey) v) Number of projects that are able to continue using AI tools upon completion of the AI Clinic. vi) Number of secondary projects arising from this consultation. vii) Number of new scientific publications that include work achieved thanks or with the support of the AI clinic. In addition, we will also measure community-related metrics. i) Impressions and community feedback on the case-studies in social networks. ii) Number of interactions in the Slack channel. iii) Number of new collaborations starting in the network
Total Budget: 16038
Budget File: pdf
Affiliations: Fundació Ersilia Open Source Initiative
LMIE Carveout: Our project fully falls within this category. While Ersilia is a UK-Spanish non-profit organisation, our target beneficiaries are researchers living in LMIE, and all our tools are developed with their requirements and needs in mind. The core Ersilia team is currently working from Spain, but they do travel often to work together with collaborators across the Global South. The broader Ersilia community is composed of open source maintainers located worldwide (including Spain, United States, Colombia, Nigeria, Uganda, Kenya, Pakistan and India) and our collaborators are located currently mostly across Africa. We aim with this bi-lingual program to broaden our services to Central and South America
Team Skills: The Ersilia team has all the necessary skills to complete the project. First, we have the in-house technical expertise required to successfully provide the services offered in this AI Clinic. Ersilia’s CSO, Dr. Duran-Frigola, has dedicated his career to the development of AI methods for drug discovery and has tens of peer-reviewed publications in the field. Second, we have lived experience with the problems we are trying to solve with this project. We have been working to increase the adoption of AI and data science in LMIE for over 5 years, and we started Ersilia 3 years ago to better serve our users. During these three years, we have had the opportunity to interact with scientists from different countries and understand their needs, from which we have devised this AI Clinic. Third, Ersilia has a solid network spanning several institutions and countries, which will facilitate the dissemination of the program and encourage participation. In particular, we can leverage the Drug Discovery Grand Challenges network via our partner the H3D Foundation, and tap onto other networks we are part of (Open Life Sciences, Software Sustainability Institute, Code for Science and Society). Finally, both Ersilia’s CSO and CEO are native spanish speakers, facilitating the offering of the program in both languages, since one of the main goals is to encourage participation from spanish-speaking countries, in which we have little presence so far.
Submission Number: 132
Loading