Artificial intelligence already permeates our daily lives. Self-learning programs such as Alexa and Siri understand language and can automatically recognize the faces of our friends and family on holiday snaps. “Machine learning has also become indispensable for biomedical research,” says Uwe Ohler, a bioinformatician at the Berlin Institute for Medical Systems Biology at the Max Delbrück Center for Molecular Medicine (MDC). “We use computers to gain new insights from large volumes of data.” Self-learning programs can independently recognize patterns and regularities in the data. They can also draw conclusions, make predictions, and provide results that improve the diagnosis, treatment, and prevention of illnesses. Even 20 years ago, they were helping the Human Genome Project to decode the human genome in its entirety.
Machine learning was also the topic of this year’s Berlin Summer Meeting, which took place from September 19 to 21 at the Berlin Institute for Medical Systems Biology (BIMSB). For 11 years now, this symposium has been bringing together scientists from different disciplines so that they can engage with each other and look beyond the limits of their own fields. This year’s event was entitled “Methods, Models and Myths – From Machine Learning to Biomedical Understanding.” It welcomed mathematicians, computer scientists, physicists, biologists, physicians, pharmacologists, and biotechnologists for three days of presentations and discussions about current developments.
Telling the computer what it needs to know
The speakers, who have been invited from universities around the world, showed the wide range of ways in which machine learning can be applied in biology and medicine. Topics included computer analysis of the increasing number of texts and images available as scientific papers, medical findings, X-ray images, and MRI images. Attendees heard how the course of a disease can be analyzed using data that a patient’s health insurance chip card has been collecting for the past ten years. And of course the talks focused on how machine learning can help geneticists identify connections between genomic variants and pinpoint predispositions for specific diseases. The “personal genome” has also been discussed in this context.
“Artificial intelligence is only as good as the data that are available, though,” says Ohler. He explains that the people behind the technology must thoroughly understand the problem at hand so that they can tell the computer exactly what it needs to do. “That’s why this symposium won’t just focus on the current opportunities – it will also look at the limits of what we can currently achieve with machine learning,” says Ohler.
This has been the first time that the Berlin Summer Meeting has been held in the BIMSB building in Berlin-Mitte. The location is very close to Humboldt-Universität and to the Charité.