A database of 200 million-plus protein shapes is hypercharging biomedical research
On July 28, DeepMind, an artificial intelligence (AI) subsidiary of Alphabet, Google's parent company, published a free, online database of the likely shapes of more than 200 million proteins known to science. This has set off a virtual carnival of biomedical research, which will transform healthcare as well as research into basic science. India must join the party.
To most people, proteins are vital nutrients, sufficient quantities of which must be ingested on a daily basis to live a healthy life. This is broadly right but lacks nuance, like defining galouti kabab as food. Proteins are the building blocks of life. Collagen, which makes up the structure of your bodily cells, is a protein. Or, rather, collagen is many proteins. Antibodies are proteins. Hormones and enzymes are proteins. Proteins allow muscles to contract and move. Haemoglobin is a protein whose job is to transport oxygen around the body.
Spiking the Protein Cocktail
A key function of DNA (deoxyribonucleic acid) is to code specific proteins inside the cell. The Covid vaccine most Indians have taken, Covishield, is a DNA (contained in a harmless virus body), whose job is to ask our body cells to produce the spike protein of the coronavirus causing Covid-19. This, in turn, triggers the production of another set of proteins, the antibodies that would attack and destroy the alien presence in the body.
DNA produces an mRNA (messenger ribonucleic acid) to instruct the cell to produce the protein in question. The Pfizer and Moderna vaccines comprise the mRNA that carries the transcription command to produce the spike protein. mRNA is far less stable than DNA, and that is why these vaccines require storage at ultra-low temperatures.
DeepMind's protein database has already been used to identify a protein that would interfere with the biology of the malaria parasite, and a malaria vaccine with 70% efficacy has been rolled out.
DeepMind is the British company that produced the AI that beat the world champion in Go, an intricate board game popular in South Korea and China. This was considered a far bigger computing accomplishment than the defeat of the then-world chess champion Garry Kasparov at the hands - in a manner of speaking - of IBM's supercomputer Deep Blue. Mastery of Go demonstrated the power of machine learning (ML), in which the computer teaches itself new things from the data it accesses.
DeepMind developed AlphaFold, an AI that forecasts the likely shapes of proteins. It is not enough to know which amino acids go into the making of a particular protein. How these string together in what sequence to form what shape is vital to know how the protein would work.
So far, scientists have been using X-ray crystallography to find out the shape of proteins. This is expensive and time-consuming. Of course, AlphaFold's forecasts would need to be confirmed the old-fashioned way. But the time taken to confirm a hypothesis, already given to you, is far less than what is spent on forming alternate hypotheses and painstakingly ruling out the false ones.
The shapes of proteins will help decipher how a parasite works, or how some pathogen produces disease inside the human body. That will guide production of proteins that would counter the harmful effect.
Read more at: https://economictimes.indiatimes.com/opinion/et-commentary/a-database-of-200-million-plus-protein-shapes-is-hypercharging-biomedical-research/articleshow/94184388.cms