#BIOSPAIN2023Interview | "The quality of the Spanish healthcare system ensures a perfect knowledge base for implementing AI tools"

Artificial intelligence has emerged as one of the disruptive areas with the greatest impact on the biotechnology sector in recent years. Despite its rapid expansion, it still faces various barriers that hinder its implementation, such as the lack of knowledge or information on the part of companies.

There is no longer any doubt about how artificial intelligence is changing the world, and its presence is increasingly prominent in all aspects of our lives. The advancement of new digital technologies like artificial intelligence has repercussions in multiple fields, ranging from healthcare and environmental protection to food safety and energy efficiency. These are all challenges that biotechnology can address and enhance through the use of artificial intelligence.

This is a perspective we analyze alongside Francisco Miguel Torres, from Owkin, a company specializing in the use of artificial intelligence to identify new therapeutic targets, biomarkers, and accelerate clinical trials. Owkin is a Bronze Sponsor of BIOSPAIN, an international flagship event in the biotechnology sector that will take place in Barcelona from September 26th to 28th.

AseBio: Owkin is positioning itself as a highly relevant global player in the field of artificial intelligence. How would you define your work?

Francisco Miguel Torres. I would define it as collaborative, innovative, and impactful.

Collaborative, because we partner with academic collaborators to unlock the true potential of the information they hold and train our predictive AI models effectively.

Innovative, because we generate these models using a multimodal strategy (we don't rely solely on clinical data; we also consider genomic, imaging data, etc.) and federated learning (where data never leaves the governance of the center, enabling international collaboration while ensuring data security). In both cases, we are pioneers and global leaders.

Impactful, because we have not only developed tools that are already in commercial phases and used by our sponsors in their clinical R&D, but from the inception of our projects, our goal is always to understand the biological heterogeneity of patients to maximize the success of any therapeutic strategy.

AseBio: Owkin started operations in Spain in September 2022. How has the journey been so far?

Francisco Miguel Torres. From our perspective, it has been excellent. The quality of the Spanish healthcare system ensures a perfect knowledge base for applying AI tools. It's true that there is still much work to be done in terms of standardization, digitization, and data curation, but it's precisely in this area that Owkin solidifies one of its value propositions for Spanish centers.

During this year, we have already initiated three significant projects and are laying the groundwork for future collaborations across virtually the entire Spanish territory. We have encountered excellent healthcare and R&D professionals who have added value to our projects, and we've found transfer offices with a great collaborative spirit and future vision.

The balance for the first twelve months of activity is very positive, and now we are scaling our presence and activities. We are actively seeking new academic partners, generating new projects, and we hope to establish Owkin as a reference in the field of AI in Spain, as it already is in the USA and the rest of Europe. Especially in France, where, in collaboration with leading healthcare centers and even the government, we are deploying this technology in the field of biomedical R&D.

AseBio: Is artificial intelligence one of the digital technologies set to revolutionize healthcare?

Francisco Miguel Torres. Without a doubt. I would say it's not just set to revolutionize; it's absolutely necessary. To tackle the immense complexity of biological systems, any prior approach without AI has proven to be insufficient. Despite the significant advances made so far, AI is the tool that will allow us to address the inherent complexity of human physiology, gene regulation, or tumor heterogeneity, to name a few examples, with confidence and in a holistic manner.

We see this at Owkin. As a biologist, I never imagined that we could analyze hundreds or thousands of patients, each with hundreds of clinical characteristics, images, laboratory results, and RNA sequences in a way that meaningful conclusions and experimentally validatable interactions could be extracted. What researcher could cross-reference what is seen in a histological section with the expression of genes A, B, and Z, and the levels of some trace element in the blood to identify a risk subpopulation in a specific tumor? It's impossible. Yet, AI makes it feasible. In fact, it does.

AseBio: Multimodal artificial intelligence in precision medicine offers significant opportunities. What doors does it open?

Francisco Miguel Torres. Firstly, it provides an agnostic approach to patient complexity. That is, by using multimodal datasets to train our Machine Learning models, we aim not to introduce any bias into our analysis. If you limit yourself, for example, to clinical data alone, you can overlook relevant interactions between, for example, physiology and gene regulation.

Additionally, it allows us to generate tools trained on multiple disease features, but ultimately only require certain variables to yield results. For example, our tool MSIntuit can estimate microsatellite instability (a molecular characteristic) by analyzing tumor biopsy images. This is because the model was trained with both images and genomic data, as well as patient outcome data. It was able to connect the dots between the molecular characteristics of the tumor and its appearance in a way that no human could.

AseBio: One of the main barriers to implementing artificial intelligence in biomedicine is the lack of information and knowledge. How do you support digitalization and preparation for artificial intelligence?

Francisco Miguel Torres. As a company, we need partners who are fully digitized, FAIR-compliant, and also AI-compliant. Finding such partners today is almost impossible. Therefore, what we do is prepare and support our partners to meet these standards. Owkin requires high-quality data sets to train the AI models we generate. Once this is done, the original data lose their interest for us, but the center can capitalize on them spectacularly in their own R&D. In other words, we create tools that researchers can leverage through their participation in our projects.

Additionally, we are also engaged in important educational and awareness-building work. We collaborate with various stakeholders within each center, helping to establish a culture of trust in AI and its security standards in many legal departments and ethics committees. We share our expertise and best practices with them.

AseBio: What do you expect from your participation in BIOSPAIN?

Francisco Miguel Torres. BIOSPAIN is a unique opportunity to introduce Owkin and engage in effective conversations with potential partners. After BIOSPAIN, we want the knowledge about Owkin, what it does, and how it does it to be more widespread. Despite being a well-established success story in France, Germany, the UK, Switzerland, and the USA, we are still a relatively unknown company in our country.

Additionally, of course, our goal is to lay the groundwork for future projects with as many Spanish innovation departments as possible. Through these initial contacts, subsequent discussions with researchers flow more smoothly, and projects are generated organically.