Ali Ertürk and Team’s AI Study Revealing Whole-Body Cancer Metastases and Drug Targets

Prof. Dr. Ali Ertürk and his team's detailed 3D metastasis imaging and targeting using AI deep learning appeared on the cover of Cell in December 2019, attracting major attention. A Bilkent Molecular Biology and Genetics graduate in 2003, he worked at Max Planck, Genentech, and Ludwig Maximillian's University. A Turkish neuroscientist, inventor, and artist in Munich, he has directed Helmholtz's Institute for Tissue Engineering and Regenerative Medicine since July 2019. Earlier work focused on transparent organs with 3-dimensional printing. The new imaging method called vDISCO shows organs transparently. Their latest work images metastases and reveals antibody targets through deep learning.

Summary of the Study Showing That Prof. Dr. Ali Ertürk and His Team Imaged Even a “Single” Cancer Cell in a Mouse Model:

Despite advances, cancer remains a major killer. More than % 90 of patients die of metastases rather than primary tumors. Drugs must reach and eliminate multiple metastases, often failing. This limitation is generally discovered during clinical trials during this process. Until now, researchers had no methods capable of visualizing the tumor-targeting ability of drugs with sufficient sensitivity and resolution to identify the most promising drug candidates during preclinical development (studies before human trials). For example, magnetic resonance imaging and bioluminescence detect large tumors but miss hundreds of potentially fatal small metastases and cannot assess whether drugs reach all. The team developed DeepMACT. DeepMACT detects even tiny metastases throughout a mouse using deep learning (multilayer artificial neural networks) technology.

DeepMACT tracks which metastases a drug targets and which it misses.

How Does DeepMACT Work?

It means deep-learning metastasis analysis in transparent tissue and has two steps: 1. vDISCO makes the whole mouse transparent. Imaging in 3 dimensions reveals all metastases and even single disseminated cells. 2. New algorithms detect metastases and analyze drug targeting. The team identified hundreds missed by other methods, with DeepMACT showing precisely which were targeted or missed. These analyses showed that highly effective antibody-based drugs may miss 30 percent of small metastases. Using this new technology, metastatic spread was analyzed in mouse models of lung, breast and pancreatic cancer. It was also possible to observe how metastases spread through the body at different time points. Today, only five percent of new drug candidates in oncology are successful. Applying DeepMACT in preclinical model systems has the potential to improve the identification and development of better drug candidates for clinical trials to an unprecedented extent. DeepMACT may therefore help significantly increase the success rate of cancer drug candidates and potentially save many lives.

Below is the study's video summary (Turkish subtitles available).

Earlier imaging was often 2-dimensional; 3-dimensional imaging is harder but more informative. Combining transparent vDISCO tissue with DeepMACT 3-dimensional metastasis imaging integrated multiple technologies. Not yet used in human trials, these methods may underpin future biotechnological innovations. In conclusion, this matters for imaging metastases responsible for % 90 of cancer deaths and revealing antibody targets, but limitations remain.

• It used mouse models injected with human cells. Human testing and clinical validation must precede use.

• Tumors contain differing cell populations. Particularly, the tumor microenvironmentis highly complex in humans and difficult to recreate in mice. More research is needed for clinical translation.