In recent years, Foundation Models have emerged as powerful tools for uncovering patterns, relationships, and the intricate statistical properties of biological molecules such proteins, DNA and RNA. By leveraging these models, researchers are poised to revolutionize DNA/RNA research, ushering in a new era of discovery and understanding in this field of biology with potential for the development of novel therapeutics and diagnostic tools. The LLM for Genomics focus group brings together SIB Members interested in RNA/DNA Foundation Models to share and discuss the latest developments, best practices, and challenges in the field, with the goal of advancing the science and application of these transformative models.

About the SIB focus groups

The focus groups aim to foster knowledge exchanges and collaborations in the community of 900 SIB Members, around specific scientific and/or transversal topics, from single-cell sequencing to equality, diversity and inclusion. View all focus groups

Large Language Models For RNA/DNA

Foundation models represent a significant breakthrough in deep learning, featuring large architectures trained on vast amounts of unlabelled data to create general-purpose representations capable of solving a large variety of down-stream tasks. Large Language Models (LLMs) for natural language, such as ChatGPT, serve as prime examples of foundation models. By using self-attention mechanisms they uncover patterns in human language, capturing its essence and underlying thought processes.

About the SIB focus groups

The focus groups aim to foster knowledge exchanges and collaborations in the community of 900 SIB Members, around specific scientific and/or transversal topics, from single-cell sequencing to equality, diversity and inclusion. View all focus groups

In biology, similar models are showing great promise, particularly in predicting protein structures and functions. Unlike convolutional neural networks, which excel at recognizing local patterns, foundation models stand out for their ability to capture long-range dependencies and complex structural arrangements. Recently, models for DNA,RNA have emerged. RNA foundation models, for example, are set to revolutionize RNA research by uncovering patterns and relationships that underlie secondary and tertiary structures, as well as regulatory functions, ushering in a new era of discovery in biology. However, several challenges must be addressed before this technology can become a research standard, including deciphering what these models truly learn, understanding their limitations, using them effectively, and continually improving their capabilities.

Goals of the focus group

The LLM for genomics focus group unites SIB Members interested in foundation models for genomics to share and discuss the latest developments, best practices, and challenges in the field. With the rapid speed of advancements and the need to stay informed about transformative technologies, this group aims to foster collaboration between domain experts and computer scientists. By expanding our reach to national and international researchers, we seek to create a platform for:

  • Staying updated: As genomic foundation models evolve rapidly, it is challenging to keep pace. This group aims to ensure that SIB members and the broader community remain informed about the latest developments, maintaining an active network of diverse experts to support and train both the Swiss community and beyond.
  • Resource sharing: Given that RNA/DNA foundation models demand significant computing and energy resources, this group seeks to build a network of research groups that can share resources, making these models more accessible to the entire community and benefit the greatest number.
  • Promoting collaboration: The best development of RNA/DNA foundation models occurs through close collaboration between computer scientists and domain experts. This focus group aims to unite these communities by offering a platform for knowledge exchange through monthly in-person and online meetings, where SIB members can present their experiences, review relevant papers, and share their latest findings.

We encourage members to actively participate by sharing their experiences with the latest tools, reviewing papers, and contributing code to a public GitHub repository dedicated to RNA/DNA foundation models.

The group currently gathers 13 SIB research groups and associated members registered to the dedicated channel #LLM-omics-meetings on the SIB Slack.

Focus group coordinating members

Are you a SIB Member and interested in joining? Contact Raphaëlle Luisier or Janna Hastings to be added to the Slack channel, and to book a date for a future presentation.