The role of AI in medicine has been growing. Recently AI has been involved in the drug creation process. For a long time, drug companies have been searching for new disease–fighting medicines by using a trial-and-error process that identifies the right compounds. The whole procedure is very complex, time–consuming and has its limitation. One is a limitation of proteins that are essential for every cellular function. Pharmaceutical companies very often use proteins in aim to eliminate symptoms or diseases inside the body and for drug creation. Unfortunately, the numbers of proteins are limited. The only ones that we are familiar with can be used to create drugs.
What if AI could do it for us?
The main aim of technology usage is to generate huge numbers of proteins that have never been seen in nature. The scientists have developed a new approach based on technologies used in Open AI programs like DALL-E2, chatbots, and other search queries. The concept enables us to “conjure up” designs for new types of novel proteins. Recently, the two labs announced separate programs that will use technology to create proteins with greater precision than ever before.
Chroma
The first one is the start-up called Generate Biomedicines. They reveal the program and called it Chroma which is referred to by the authors as DALL-E 2 biology. Chroma generators could be used to create proteins with new properties like size, functions, or shape. The whole idea opens our doors to drug inventions that could fulfill all our requests. It is a huge possibility that enables scientists to discover something in a few minutes that took our evolution millions of years.
How does it work?
The concept is modeled on a DALL-E2. AI used as a diffusion model eliminates random perturbations applied to data from their source. That’s how unorganized data is arranged in a coherent piece. Similarly, Chroma may synthesize unique proteins with specific characteristics not governed by set constraints on how the product should look.
RoseTTAFold
A different approach has been introduced by Baker’s team. They invited something slightly different but with the same results. Baker’s team decided to use RoseTTAFold Diffusion which guides the overall generative process by analyzing how pieces of protein supplied by a neutral network fit together. Baker Lab believes that the proteins built by AI are a better fit for essential hormones than existing protein drugs.
A protein structure generated by RoseTTAFold Diffusion (left) and the same structure created in the lab (right)
Natural language processing
This idea is based on Open Ai’s Chatbots and their ability to generate human-like responses. They discovered the collocation between biological codes and search queries. To both of them, you need to respond by using a series of letters. Proteins are built up of amino acids, and scientists employ particular notation to record the sequences. Proteins are represented as lengthy, sentence-like combinations with each amino acid corresponding to a single letter of the alphabet. Natural language algorithms can be successfully used for creating protein-language models. The models encode the so-called ‘grammar of proteins’ to predict the sequences of new drug molecules. In consequence, the time required for drug discovery might be shortened to months.
As you can see, there are existing solutions to the problem of limited proteins. Are they sufficient to safely develop new drugs? This will be discovered in the near future. I expect that AI-powered models will be developed with more effective pharmaceuticals to treat incurable diseases and improve the quality of life for terminally ill patients and their families.
Sources :
How AI That Powers Chatbots and Search Queries Could Discover New Drugs – WSJ
Biotech labs are using AI inspired by DALL-E to invent new drugs | MIT Technology Review
Using all of the skills that Silicon Valley and the tech ecosystem have generated, the confluence of technology, drug development, and biology will result in the development of better treatments more quickly, allowing us to make an even bigger impact on patients.
But how safe is this method as we know everything that is new doesn’t mean its safe and taking into account something like this can only come with time at its use.