Luigi Vitagliano is a Research Director at the Institute of Biostructures and Bioimaging in Naples, Italy. He shares his AlphaFold story.
Being a structural biologist in the age of AlphaFold is like the early days of gold mining. Before this technology, everyone was doing painstaking work to find individual gold nuggets, cleaning them and looking at them one by one. Then, all of a sudden, a gold mine appeared. We couldn’t believe our luck.
For 30 years, I’ve been studying the proteins encoded in our DNA. Within most human cells, there are somewhere between 20,000 and 100,000 different proteins. In certain instances, the way the string of amino acids in a protein takes its shape, also known as 'protein folding' can be full of irregularities, and these are linked to lots of diseases.
Recently, I’ve been looking at a family of human proteins, known as potassium channel tetramerisation domain (KCTD) proteins, that are particularly poorly understood. What is particularly interesting about mutations in these proteins - caused by genetic mutations - is the range of diseases that they are linked to: from schizophrenia to autism, and leukaemia to colorectal cancers, as well as brain and movement disorders.
As new proteins are constantly being made inside cells, old or defective ones need to be removed. There are 25 kinds of KCTD proteins in humans, and four-fifths of them seek out other proteins and mark them for degradation and destruction. This process is called ubiquitination and it is essential for keeping cells healthy and helping to prevent disease.
When KCTD proteins don’t work properly, the consequences can be debilitating to our health. However there’s a lot we don’t understand about them, too. About one-fifth of KCTD proteins inside cells were mysteries to scientists like me: we had no idea what they do, and therefore how to prevent them mutating and causing disease. Until now, we’ve had very little structural information on them, which has been a major barrier to KCTD research.
"Entire generations of eminent scientists have been unable to solve this problem. Then, almost miraculously, a solution appeared."
- Luigi Vitagliano, Research Director
The structures predicted by AlphaFold revealed that over the course of evolution their structures have remained very similar despite having very different genetic codes. This was a significant breakthrough. Previously, we’ve relied on genetics to assess the similarities or differences between proteins. Based on genes alone, we thought these proteins would be very different.
Using AlphaFold, we were able to build a new evolutionary family tree based on the shape of these proteins rather than their genetic sequence. Evolutionary trees are usually built using genetic information, but they don’t take structural similarities into account. Structure relates to function, so using this approach is thrilling – it could reveal all kinds of mysteries about which KCTD proteins have similar functions and how these functions evolved over time.
I used AlphaFold to look at and compare the structure of all 25 KCTD proteins for similarities and differences, to identify which parts of these proteins are important. To our delight, AlphaFold's predicted structures appeared to be very accurate.
For example, we already knew that one section of the KCTD proteins – the BTB domain – was similar amongst all family members, and so we presumed this was the most important part. AlphaFold has revealed many more additional structural similarities amongst these proteins and has opened up an entirely new realm of exploration.
For 60 years – including the 30 years that I’ve been working in this field – we’ve tried and failed to find the connection between sequences and structures. Entire generations of eminent scientists have been unable to solve this problem. Then, almost miraculously, this solution appeared. All of our data, the structural information for all members of the KCTD family, has come from AlphaFold. Without it, this study couldn’t have been done at all.
My feeling was that AlphaFold was a dream. If somebody had told me that in two years we will have over 200 million protein structures, I wouldn’t have believed them. Now, what lies in the decades ahead is finding out exactly what these proteins do. There’s a lot more excitement and discovery ahead.