Immunotherapy, which involves activating the patient’s immune system to fight a particular disease, is one of the fastest growing branches of medicine. Over the next 10 years, the market for antibodies used in this field could triple, says analysts at Fact.MR. Polish start-up Genotic is leveraging artificial intelligence in its work on antibodies targeted at specific antigens, significantly reducing the process from several months to just 21 days.
“Artificial intelligence provides the opportunity to automate many processes in biotechnology and pharmacy, starting with the search for new drug candidates, not only small molecules, but the much more difficult, namely antibodies. Biological drugs, i.e. antibodies, are among the best drugs, so they are much more difficult to design, but deep learning networks allow us to create even personalized antibodies for specific epitopes of targets in patients in a very short time. This is a huge revolution similar to the scale of the discovery of electricity, which is currently taking place in the world of biotechnology and pharmacy,” says Grzegorz Warzecha, founder of Genotic.
Using artificial intelligence, the Polish start-up designs, tests and produces antibodies that can be used in the diagnosis and treatment of infectious, autoimmune and oncological diseases. Protein structures are predicted based on sequences. This greatly streamlines the process and the antibodies targeting a specific antigen.
“We primarily use artificial intelligence to design personalized antibodies highly specific to specific targets, such as HER2 in the case of breast cancer. We want to emerge with a large base of antibodies for many different targets in the R&D market and diagnostics. At the same time, we are ambitiously working on automating processes in laboratories using deep learning networks,” says Warzecha.
An example of using artificial intelligence in medicine is the AlphaFold 3 model, developed by Google DeepMind and Isomorphic Labs. The solution models large biomolecules, such as proteins, DNA and RNA, as well as small molecules. It can also model the chemical modifications of these molecules that disrupt cell function and lead to disease. In practice, this represents a revolutionary change in the ability to develop new, innovative therapies. Genotic used AlphaFold to develop its platform for generating antibody structures.
“The greatest value of deep learning networks is the ability to perform drug design processes entirely digitally on a computer, known as in silico, because it completely changes and simplifies the process of finding new candidates. We can simplify a lot of laboratory work by finding good candidates early on using deep learning networks,” explains the Genotic expert.
The basic technology in antibody work is their design in an animal’s body. For example, in the case of a virus, the isolated virus is injected into the animal, and after long months of waiting, the antibodies formed are isolated from the animal’s blood – if an immune reaction has occurred. Only based on this is the sequencing of the protein and the determination of the structure carried out, and only after this process can production start.
“We are able to design an antibody within 48 hours on 200-300 graphics cards, which means that in literally two days we have many verified candidates who can be produced and passed on to laboratory verification,” assesses Warzecha.
Next comes checking the effectiveness and specificity of candidate antibodies, and ultimately, from thousands of nominated candidates, those with the best parameters are selected in the laboratory.
Monoclonal antibodies are already the most desired therapeutic element, for example in hematological diseases: multiple myeloma or DLBC lymphoma. Dual antibodies on the one hand recognize abnormal cells, and on the other – teach the immune system how to fight them. According to Fact.MR, the global antibody market exceeded $197 billion in 2022. By 2032, revenues will increase more than threefold – up to $608 billion.
“I believe that our future as patients will also involve companies creating on-demand drugs for us, sequencing our RNA-seq, DNA, i.e. seeing the structures of our disease, will create personalized drugs in a very short time and deliver them to the patient in three weeks from the start of the tests,” predicts Genotic’s founder. “It’s a new trend that’s coming up worldwide. Deep learning networks are good at mapping representations, i.e. creating certain structures, such as protein molecules. The progress over the past two years has been radical.
