group of people, waving their hands

AI’s adventures in genomics

This year’s Berlin Institute of Systems Biology of the Max Delbrück Center summer meeting explored “AI’s Adventures in Genomics.” The meeting provided molecular and computational biologists a forum to discuss the latest advances, challenges, and prospects in harnessing AI for biomedical research.

How can researchers harness the potential of AI to extract meaning out of terabytes of genomic data? That was the subject of the 17th annual Summer Meeting at the Berlin Institute of Systems Biology of the Max Delbrück Center (MDC-BIMSB). The meeting, “AI’s Adventures in Genomics,” brought together molecular and computational biologists to discuss how researchers are using AI tools to help analyze the reams of data generated from molecular biology experiments.

It has been more than 20 years since researchers completed a draft sequence of the human genome in 2000. At the time, the feat was a landmark achievement that some likened to man landing on the moon. But it soon became apparent that turning that data into knowledge would likely require a similar Herculean effort. In 2003, molecular biologist Eric Lander and principal leader of the international Human Genome Project famously quipped: Genome: bought the book, hard to read.

Two decades later and armed with increasingly powerful AI tools, computational biologists are experimenting with machine learning tools to develop models that can crunch data on a scale that has previously not been possible.

Applying AI to “omic” data

At the meeting, Dr. Julien Gagneur, Professor of Computational Molecular Medicine at Technical University, Munich, for example, demonstrated how he trained an algorithm to find functional sequences in the non-coding regions of the yeast genome. He showed that a model can start learning the features and connections in data without being instructed to do so.

Dr. Andreas Keller, Professor and Chair for Clinical Bioinformatics at Saarland University, University Hospital, discussed how he uses machine learning methods to understand the molecular processes that cause aging. He presented single-cell RNA sequencing data from the blood of Alzheimer’s patients, which he called a “complex landscape of data.”

Keller found significant differences in the gene expression patterns of men and women in the blood of people with the disease, but offered no explanation for what the differences might mean. “It’s just an observation,” he said, adding “my brain can’t capture the complexity.But he sounded confident that “AI will help us solve these issues.”

Modeling cellular processes

Other researchers, such as Dr. Rebekka Burkholz, Group Leader of the Relational Machine Learning Lab at CISPA Helmholtz Center for Information Security and Dr. Joshua Welch, Associate Professor of Computational Medicine and Bioinformatics at the University of Michigan, discussed developing computational models to understand the timing of specific molecular events in cells: Burkholz is studying the order in which mutations in cancer cells arise. Welch shared his experience using MultiVelo, a tool for modeling cell fate transitions over time from single-cell multi-omic data.

MDC-BIMSB’s own Daniel León Perinán, a doctoral student in the Systems Biology of Gene Regulatory Elements Lab of Professor Nikolaus Rajewsky, showed how he used Open-STa sequencing-based, open-source experimental and computational resourceto reconstruct gene expression in cells within tissues in 3D, with subcellular precision.

AI to design better therapeutics

The second day of the conference featured several talks on how researchers are leveraging AI to design novel therapeutics. Dr. Ewa Szczurek, Co-Director of the Institute for AI for Health at Helmholtz Munich showed how HydrAMP, a machine learning tool developed in her lab, could design antimicrobial peptides – alternatives to conventional antibiotics and that are effective against antibiotic resistant pathogens. Lab experiments on five bacterial strains confirmed that the peptides could effectively kill bacteria. Later in the day, Professor Georg Seelig at the University of Washington in Seattle discussed how machine-learning tools can guide sequence design for mRNA and gene therapy applications.

The presentations were very diverse and showed “how AI can speed up basic science, our work with big data, all the way to research on therapeutics,” said Professor Uwe Ohler, Group Leader of the Computational and Regulatory Genomics Lab at MDC-BIMSB. “We have already started discussions with several speakers about collaborating on joint projects to further develop these tools.”

Text: Gunjan Sinha


Further information