accelerated vaccine development

A groundbreaking announcement of 2024 was the Nobel Prize in chemistry being awarded to two AI researchers from Google DeepMind along with a biochemist working at the University of Washington, for their work in developing an AI model called AlphaFold which can have a disruptive impact on the healthcare industry moving forward. This announcement is a strong testament to the wave of transformation that AI is set to bring forth which I had highlighted in a previous article I penned - futuristic realm of AI. So, what exactly is it that the Nobel Prize winners uncovered and how does it leverage the capabilities of AI to solve a very crucial problem in the healthcare industry? Before I get into it, I would like to dive into the process of conventional vaccine or drug discovery. Drug discovery requires understanding the spatial biochemical structure of proteins that constitutes the infective virus against which the drug is being developed. All living organisms from bacteria and viruses to animals and plants ultimately consists of special sequences of certain molecules called amino acids which make up various protein molecules that are the fundamental building blocks of living organisms. The amino acid sequences are developed through the help of special biological instructions i.e. genetic code (contained in DNA or RNA sequences) that get executed by biological (computer-like) entities called ribosomes to produce the various life-possessing molecular structures we see that truly differentiate the living from the non-living which I also covered in depth in another article I penned - dissecting genetic code. Now coming back to the vaccine discovery process, the sequence of protein molecules of a virus fold into unique 3D shapes and being able to get a spatial rendering of the folded 3D shapes is essential to understand how the virus infects us and how to in turn combat the virus through external intervention. Let us look at how a virus infects humans by taking an example of the (unfortunately) popular COVID 19 virus, the spike protein of COVID 19 virus can attach to a receptor on the surface of human cells like how a key fits into a lock depending on the folded structure of the virus proteins. After the virus binds to the human cells, the viral genome (i.e. genetic code of the virus) gets transmitted into the DNA (i.e. genetic code of human) of human cells which ends up altering essential biological instructions stored in those cells which consequentially affects the essential biological functions that it performed as a result of those biological instructions. Similar to how a virus attaches to a human cell and infects it, external treatment through vaccines works by attaching a treatment molecule onto the virus that can alter its genetic code and thereby inhibit its function. Vaccine development is the process of finding the appropriate treatment molecule. Vaccine development involves few key steps: firstly, we need to derive the sequence of virus proteins through mRNA sequences (i.e. genetic code that produces those sequence of protein molecules). Once we derive the sequence of the protein molecules in the virus, we then need to derive the 3D structure of the protein sequence that is formed as a result of protein folding. From the 3D structure, we need to derive the target of the virus protein that binds to the human cells thereby causing the infection. Then based on knowledge of the target, we generate the best possible treatment molecule that can inhibit the malicious function of the target on the virus. The steps of identifying the 3D structure from the protein sequence, target on the 3D structure and treatment molecules are extremely time consuming and requires extensive trial-and-error processes. Now coming to the Nobel Prize winning discovery by the DeepMind AI researchers- AlphaFold, it is an AI model which aids in the step of identifying the 3D structure that the sequence of protein molecules fold into. DeepMind’s AlphaFold is an AI model that is trained on a large database of previously discovered 3D protein structures (output) of corresponding protein sequences (input) which gives an objective mathematical function that can take in new inputs (i.e. protein sequences) and predict the requisite output (i.e. the protein 3D structure) with a high level of accuracy. The process of easily identifying how different protein sequences fold into different 3D structures has traditionally been an extremely challenging task that has plagued biological scientists for a long time and has been termed as the ‘protein folding problem’. Hence, the AlphaFold discovery has been heralded by the biological community as having solved a ‘fifty-ear-old grand challenge in biology’ which can now help significantly accelerate the vaccine development process. AI can be further leveraged in the steps that follow identifying the protein 3D structure identification process. AI models can be used to propose targets on the 3D structure on which treatment molecule gets attached and given the target, AI models can narrow the search for treatment molecules (i.e. vaccine) by training them on a large database of past data similar to how AlphaFold model was trained. Intervention of AI in healthcare (as evidenced by DeepMind’s AlphaFold) seems like that start of a new paradigm where computer algorithms can be seamlessly interweaved into the inner functioning of biological organisms facilitating a growingly harmonious and deeper symbiosis between machines and humans.