Fighting cancer with AI

Every year, around half a million people are diagnosed with cancer – and the number is rising. Yet ever-improving diagnostic options clearly show that cancer is a highly individual disease, and that every patient requires a therapy that is tailored to his or her specific needs. In an effort to achieve this, scientists at the Berlin Institute of Health (BIH) are enlisting the help of artificial intelligence (AI) in various projects. With it, they are able to characterize individual cancer cells in great detail, select the appropriate drugs for patients, and develop a medical file specifically for cancer patients that can record all disease and treatment data and accompany them on their often long journey with the disease.

Professor Angelika Eggert is Director of the Department of Pediatrics at Charité with a focus on oncology and hematology, and spokesperson of the BIH’s Multi-Omics research platform: “Using new molecular technologies, we are already able to characterize tumors very precisely – right down to the level of individual cells.” Such technologies enable the entire genome of these cells to be decoded, all proteins to be characterized, the metabolic processes of the cells to be followed, and biomarkers or surface molecules to be discovered. This, of course, results in huge amounts of data. “In order to record this data and draw the correct conclusions from it, we also rely on artificial intelligence,” says Eggert.

The analysis of this molecular tumor data provides information on possible target molecules for therapy. Scientists can also calculate the patient’s individual risk of relapse and adjust the aggressiveness of the treatment accordingly. “We may also try to offer patients new experimental therapies, especially when the risk is very high,” says Eggert. But this complex procedure is not suitable for every cancer diagnosis: “This is for patients whom we have not yet been able to help effectively, and for whom we therefore have to make additional efforts – including with the help of artificial intelligence.”

Growing organoids from cancer stem cells

Led by biologist Dr. Christian Conrad, scientists at the BIH Digital Health Center are examining primary tissue samples of surgically removed tumors. “These are often only very small amounts – not enough for us to test the effectiveness of 50 different drugs,” explains Conrad. “We therefore use the stem cells contained in this tissue to cultivate small, three-dimensional tissue pieces known as organoids, and use these to carry out the tests. Under the microscope, we can observe whether a substance has caused the tumor to shrink, whether it kills tumor cells, or whether it has no effect.”

Because there are so many samples to be tested, the researchers are working on automating the procedure. “We use culture plates that can hold up to 100 organoids at a time,” says Conrad. ”These are automatically analyzed under the microscope and the data is evaluated directly in the computer. Our goal is to develop models that can provide information about which drug has been most effective in fighting which patient tumor.”

In parallel, the scientists are also analyzing the genetic expression of these mini-tumors. This provides them with insight into the genetic changes contained in the tumor and allows them to predict which targeted drugs are most likely to be active in combating the tumor cells. “In most cases, our prediction based on the sequencing data corresponds to the observations made under the microscope, but sometimes it does not,” reports Conrad. ”We want to understand the relationship between morphology – i.e., the appearance of the organoids – and gene expression.”

“Our goal is to be able to offer various cancer patients the most effective therapy with as few side effects as possible,” summarizes Professor Roland Eils, founding director of the BIH Digital Health Center. “In the future, we also want to further develop the method so that we can use it for as many types of cancer as possible.”

An electronic medical record for cancer patients

Professor Sylvia Thun heads the BIH Core Unit eHealth and Interoperability. She wants to develop an electronic medical record for cancer patients based on international IT standards (FHIR, SNOMED, LOINC). “These patients are often embarking on a long journey through different doctors and hospitals,” explains Thun. “First, they are diagnosed by their GP, then they go into the hospital where they undergo further examinations, surgery and radiation, possibly chemotherapy by a resident oncologist, followed by regular follow-up appointments – and then the tumor may return years later.” During this time, examination and treatment results accumulate that can fill several folders. “We want to ease the burden on patients and enable them to bring together all their personal medical history data in one electronic patient file,” says Thun.

The problem is that the data gained from laboratory diagnostics, pathology, imaging, tissue examinations and gene sequencing all have different formats, the course of treatment and the drugs taken may not be documented or may be documented in different languages, and how the patient actually feels throughout the therapy would have to be recorded by the individual. “In order to be able to use the data for further processing or for research, however, it must be available in a format that is structured and able to be read ideally by both humans and machines,” explains Thun. “Our goal is to collect all disease data in a standardized format and to structure it in such a way that cancer patients can keep their personal data with them in their electronic patient file and make it available to their doctor or to researchers whenever and however they want.” Together with the help of artificial intelligence and international networks that communicate with each other in standardized languages, such files could help determine the best therapy for individual patients. Furthermore, by combining and evaluating data from many cancer patients, completely new relationships may also be revealed that could lead to new forms of treatment.