Project "ikarus" provides new insights for cancer research

Artificial intelligence (AI) is about to become a game changer, especially in diagnostics. However, there are still limits to the use of AI. Though a huge amount of data sets already exists, it is still not that easy to use them. Dr. Altuna Akalin had to recognize this as well. The head of the Max Delbrück Center's (MDC) technology platform for "Bioinformatics and Omics Data Science" developed "ikarus" with his team. The machine learning algorithm can distinguish between cancerous and healthy cells.

In an interview with MEDICA-tradefair.com, he reported on the difficulties he and his team had to overcome and how "ikarus" has already contributed to gain completely new insights. 

Dr. Akalin, you have just developed "ikarus", a machine learning program that can distinguish cancer cells from healthy cells. How does it work?

Dr. Altuna Akalin: We developed a software that works in single-cell resolution and finds gene expression patterns to specific cell types. In case of cancer, the method found a pattern in tumor cells that is common to different types of cancer, consisting of a characteristic combination of genes. The algorithm also detected types of genes in the pattern that had never been clearly linked to cancer before.



What was the biggest challenge in developing ikarus?

Dr. Akalin: Single-cell sequencing data sets are often noisy. That means the information they contain about the molecular characteristics of individual cells is not very precise – perhaps because a different number of genes is detected in each cell, or because the samples are not always processed the same way. Our team sifted through countless publications and contacted quite a few research groups in order to get adequate data sets. The team ultimately used data from lung and colorectal cancer cells to train the algorithm before applying it to data sets of other kinds of tumors.

What are the current limitations or difficulties of Data Science in general?

Dr. Akalin: The general difficulty in data science is to find data sets and cleaning them up. In life sciences, this problem is magnified because it is hard to get relevant datasets. One has to create them from scratch or make a long curation process. In either case, the data will be noisy due to biological and technical variation.

You want to generalize the "ikarus" approach so that it can distinguish all possible cell types in a biopsy, not just tumor cells. What applications would be conceivable then?

Dr. Akalin: Our project aims to go far beyond the classification of "healthy" versus "cancerous" cells. In initial tests, ikarus already demonstrated that the method can also distinguish other types (and certain subtypes) of cells from tumor cells. We want to make the approach more comprehensive by developing it further so that it can distinguish between all possible cell types in a biopsy.

In hospitals, pathologists tend only to examine tissue samples of tumors under the microscope in order to identify the various cell types. It is laborious, time-consuming work. With ikarus, this step could one day become a fully automated process. Furthermore, the data could be used to draw conclusions about the tumor’s immediate environment. And that could help doctors to choose the best therapy, for the makeup of the cancerous tissue and the microenvironment often indicate whether a certain treatment or medication will be effective or not.

How is Big Data changing medicine right now?

Dr. Akalin: More diagnostic approaches are initially mined from big datasets. Recently, more and more companies are using large datasets on topics ranging from target identification to drug discovery. New drugs and diagnostics are more likely to be developed through analysis of large data sets.