Background and aim of the project

The increasing availability of large and complex data sets has also turned the life sciences, human genomics, and their application in (bio)medicine into attractive application areas for artificial intelligence (AI) in the form of advanced machine learning (ML) methods, such as deep learning (DL). Sequence data of the human genome as well as data on genes and proteins that are expressed in various body cells and organs play a decisive role in this context. Similarly, data on the functions of genome sequences derived from new genome editing techniques (such as CRISPR/Cas) and the targeted genome-wide modifications they enable are becoming increasingly important. In addition, there is an increasing amount of clinical data on diseases and their courses from biobanks or electronic patient records (e.g., medical image data or blood values and other biomarkers) that can be combined with the genome data.

The aim of the project was to conduct an interdisciplinary analysis of the increasing convergence of developments in AI and human genomics as well as the innovation system that promotes and shapes them. One goal was to look at both the nature or (new) quality of the knowledge that can actually be obtained through these developments and its implications for possible applications. This approach should also shed light on realistic applications in the short to medium term, with their opportunities and specific ethical, social, economic, and regulatory challenges. Furthermore, the analysis of these opportunities and challenges should form the basis for identifying areas for political action as well as recommendations for such action, with a particular focus on research and innovation policy.

Summary of results

A. Development of the field

The highly dynamic development of publication and patent activities at the interface of AI and human genomics worldwide over the last six or seven years can be taken as an indication of the emergence of a new innovation system. In relation to the respective activities in the fields of AI and human genomics as a whole, the combination of AI and human genomics appears to (still) be a niche. However, if the observed dynamics continue, research and development (R&D) in human genomics will already be strongly influenced by AI in the coming decade. Major players in the field currently include – in addition to large international academic and clinical research consortia – venture capital and large IT firms, and increasingly large pharmaceutical companies, which often partner with innovative start-up companies.

B. Potentials and expectations

Potentials and associated expectations have been identified via evidence mapping from publication and patent data, first, in basic research. There, first AI/ML-based methods for rapid and comprehensive (i.e., genome-wide) prediction of causal genetic changes, especially for complex diseases (that are typically dependent on multiple genetic changes or genes and environmental factors) have been developed. These methods can help identify putative causes of disease and pathophysiological mechanisms. Second, such knowledge is used in combination with other AI-based techniques for more targeted and rapid drug development or drug repurposing. Finally, new diagnostic and prognostic methods for diseases, courses of therapies or disease risks are becoming possible. These include, for example, the diagnosis of rare genetic diseases or approaches for the early detection of cancer as well as for predictions or adaptations of corresponding therapies by analyzing minute amounts of DNA in the blood (“liquid biopsies”).

C. Challenges and fields of action

Future development is associated with a number of challenges that lie not least within the sphere of public research and innovation policy. These include, in particular, the generation of large and high-quality genome and health data and their exchange and accessibility across borders and different regulatory conditions, the establishment or restructuring of corresponding research and data infrastructures, the promotion of research and start-up companies, as well as socio-cultural aspects such as acceptance on the part of users and the public, and the consideration of social norms and values.

D. Implications for research and innovation policy

In view of these challenges and the dynamic development of the field, especially in China and the USA, a stakeholder-based research and innovation policy strategy process should be set up. Such a strategy process should focus on how Germany and Europe can strengthen their position in the responsible and value-oriented development of AI innovations in genomic medicine and their attractiveness for international cooperation, talent and investment. Building on this, the process should also work toward the implementation of measures that go beyond the research and innovation policy activities already underway or planned. There is a particular need for action in the experimental validation and clinical evaluation of AI-based predictions, the initiation of large and diverse genome data and biobanks, the development of concrete guidelines for academic researchers and companies on regulatory compliance with complex national or supranational regulations, and the initiation of the broadest possible societal discourse with the systematic participation of diverse societal actors and the public.

• Policy Brief “Künstliche Intelligenz in der genomischen Medizin – Potenziale und Handlungsbedarf” (PDF)
• Evidence maps “Artificial intelligence in human genomics” (PDF)
• Journal article “AI models and the future of genomic research and medicine: True sons of knowledge?”

Tel.: +49 721 608-23293
E-mail