Feature: Another dimension in protein interaction

Sitting in a darkened room, I gaze up at a giant projection of what could be an abstract artwork, or perhaps an advanced computer game. Arrayed before me is a constellation of tiny floating dots, some loosely scattered, some clustered within larger structures, and many of them bound together by a startling array of glowing lines looking for all the world like a giant space battle. But this is not art (though it’s clearly artistic), nor is this a game; this is an up close and personal view a common yeast cell.

“Imagine for a moment that you’ve come up to a yeast cell and you’ve just poked your head inside the outside membrane,” says Professor Marc Wilkins, the ‘father’ of proteomics and Director of the NSW Systems Biology Initiative, School of Biotechnology and Biomolecular Sciences at the University of New South Wales. Wilkins is the project lead for the technology on display, a new three dimensional visualisation tool for exploring the inner workings of cells, called The Interactorium.

Wilkins points out the various organelles within the cell – there’s the nucleus, there’s the mitochondria – and all the thousands of dots scattered around, those are the proteins. It’s all there. Then, with a couple of clicks of the mouse, Yose Widjaja – PhD student and the brains behind the coding of The Interactorium – gracefully rotates the mass of dots and lines, selects a single ‘node’, and the view zooms in to reveal the protein in detail.

Mere words and still images simply can’t do justice to the beauty and elegance of The Interactorium. For a cell – even one as ‘simple’ as a yeast cell – is a startlingly complex thing; thousands of individual proteins, each with potentially dozens of interactions. Trying to visualise all this through with a conventional database, or even a two dimensional diagram, rapidly becomes an exercise in folly. But with The Interactorium it’s all there in glorious 3D, intuitively navigable, displaying a vast amount of information in an immediately digestible form.

The Interactorium is also more than just a tool for generating pretty 3D images. Wilkins hopes it will not only enhance our understanding of the inner workings and dynamics of cells – healthy and diseased – but also aid in the development of new drugs that target specific proteins with super-surgical precision.

---PB---

Information overload The trick has been how best to represent the multitude of interactions that take place inside a cell in an intelligible way. Even with a yeast cell you’re staring down the barrel of 6,000 genes, let alone the 22,000-odd genes for a human cell. And then you have the host of interactions between them all. Vast databases have been constructed containing information about each of these genes and proteins, but just staring at a table of data doesn’t help much when it comes to trying to make sense of it all.

However, we already have a faculty that is designed specifically to process a vast amount of information rapidly and effortlessly: vision. By making the leap from data to an image, we can immediately see patterns and make inferences that would have been nigh impossible just by pouring over data. Just think about how much easier it is to read a graph rather than look at a table in a spreadsheet.

Visualisation tools for protein interactions aren’t a new invention. Cytoscape and VisANT, for example, have both been around for a number of years, and have proved useful in representing different parts of cells. But both are limited to two dimensional views, which can easily become overwhelmed by the data. Even 3D tools like GEOMI (Geometry for Maximum Insight), which was developed by NICTA, struggles to handle even a relatively simple cell, like yeast.

“One of the challenges we’ve always had with the technology available to us is if we wanted to move up to a very large number of proteins and a large number of interactions, the software available to us didn’t support that,” says Wilkins. “You’d reach a certain level whereby the existing software hits the wall around about a thousand proteins.”

Even with the limited scope of older visualisation tools, they also proved slow and cumbersome to use. “One thing which we’d be the very first to complain about with our old technology was the navigability in three dimensions was pretty slow. It was a Java-based platform and it really wasn’t going to be as fast and powerful for being able to build things in three dimensions and navigate them in the kind of way we can do here now.”

This is where Yose Widjaja steped in. Last year he was an Honours student working with visualisation expert, Dr Tim Lambert, in the School of Computer Science and Engineering. Widjaja’s Honours project was a visualisation tool called Skyrails, which was originally intended to model things like social networks. However, with a deft piece of interdepartmental collaboration, he soon discovered another potential application for Skyrails: the interactome.

“It was an astonishingly fast and powerful platform for generating a navigable three dimensional space,” says Wilkins. “We got talking about some of the types of things we were doing and this collaboration arose from that in which we have now built a virtual cell.”

Thus was The Interactorium born, and it now forms the crux of Widjaja’s PhD project, which he began this year.

---PB---

Cell space

When you first fire up The Interactorium – which you can do on just about any recent desktop or laptop computer with a 3D graphics card – you can choose to view the cell in one of two ways: the first is the Complex Viewer, which just shows the interactome; the second is the Virtual Cell, which also adds localisation data.

In the latter view, you’re treated to a wireframe panorama of the guts of the cell, organelles and all. “As you navigate around, the size of the organelles bear some approximation to the number of proteins you actually see in each of them,” says Wilkins. “We haven’t literally built these organelles to be the size and shape of an organelle – of course that can be done if that’s imperative.

“There seems to be a lot of space that contains nothing, but we can then show all the proteins that are in the cytoplasm, and all of a sudden we see that the cytoplasm is a very busy place. All the lines are the interactions. That’s when you start to see all of the proteins and the protein complexes. We like to look at the cell as not just as set of individual proteins but protein complexes – we think it’s the biologically appropriate view. It’s a slightly higher-level view than you see in a lot of networks. A lot of networks have one dot is one protein or one dot is one gene. We instead have very much focused on what is the functional unit inside the cell and sought to build it up from there.”

But the crucial part is visualising the interactions. “The lines between the proteins show the interactions, but also what’s known as the ‘quality’ of the interaction, measured by different methods.” Wilkins is referring to the fact that some lines connecting two different proteins are visibly thicker than others.

“Beneath this there’s a very sophisticated database about what the proteins are, where they’re found inside the cell and what complexes they’re known to be part of. Mousing over the lines between two proteins says whether there’s evidence in the literature that support this interaction.”

So at a glance, you can immediately get an impression of the confidence in the interaction or the complex being displayed. “It makes it facile to see straight away what the quality of the data is. You look at that and you get the feeling straight away that this is a well defined complex. Rather than having to look at an Excel spreadsheet and having to work through stacks of different papers and publicational databases, that type of thing is a very much at your fingertips straight away.”

This is all powered by a sophisticated database – in this case compiled by Simone Li, who works with Wilkins at the Systems Biology Initiative at UNSW – which brings together a broad selection of the literature on the yeast cell. In fact, one of the strengths of The Interactorium is that none of the data about the cell is hard coded. Plug in a different database – say for a human cell – and The Interactorium will lay it all out for you.

They’ve even go so far as to add a deeper layer into the visualisation. “We’ve got a virtual cell, we’ve got organelles, we’ve got complexes, we’ve got proteins – if we keep on working through that hierarchy the next logical thing is if there’s structural data for particular proteins. So when you get sufficiently close to a particular node of interest, it’ll render the protein structure in three dimensions for you.”

---PB---

Predictably powerful

One of the applications of The Interactorium that interests Wilkins and his team is the ability to get a better understanding of the inner workings of a cell, such as which elements are static and which are dynamic.

However, beyond displaying what we know about a cell, The Interactorium can also give us an idea of the blanks, of what we don’t know. “So if you have a group of five proteins that all interact with each other in this network, and the function of four of them are known, you can make extremely strong hypotheses and predictions of what the function of the next one is. And, of course, functional prediction is a very important thing to be able to do in the biosciences.

“In the human genome there’s still around about 11,000 genes for which we have no known function. Even in yeast – the humble little yeast – with only 6,200 genes, I think there’s still a good 1,000 or so genes for which there’s no function. So it’s guilt by association you can do inside networks, which is always kind of neat.”

Wilkins has also had interest in The Interactorium from researchers working in a range of fields, from cancer to drug discovery. “There’s a lot of interest in terms of using network-based analysis for understanding changes in cells, whether they are stem cells as they are going from pluripotency to a final tissue type or looking at changes in cancer cells.

“People are also interested in using this kind of thing for drug discovery: understanding networks, understanding pathways, seeing how those are associated with phenotypes or the genesis of disease. If there are particular proteins that have a sensitive interaction inside these networks, you might ask the question: ‘can we block that with a particular drug or peptide.’”

The Interactorium might also have a use in predicting side effects of particular drugs. “If you make a drug against a particular protein, is it only going to be acting against one protein or one interaction? And if you block that interaction, in a network sense, is the cell going to be able to do the same thing by another path?”

It’s early days for The Interactorium – the first paper on it should appear in Proteomics by the time you read this – but already its potential to reveal secrets of protein interaction are plain to see.