Model citizens

Over the years, I have had the opportunity to interview and write about a lot of groups involved in modeling molecular structures and biological systems. To a person, everyone has been overwhelmingly excited by the potential insights that their models will provide researchers looking to understand human disease, develop new drugs, or engineer new enzyme activities. And because people understand pretty pictures much better than chemical conceptualizations, the enthusiasm is infectious.

 

But just as the runway models of Paris and New York are idealized (?) representations of what it is to be a human being, so too are the computer-generated models that rotate on the video display terminals of Cambridge and Heidelberg idealized representations of biomolecular realities. And like the television shows that claim to be "ripped from the headlines", so do these models exclaim "based on a true story."

 

All this to say that, while I believe in the value of modeling, I worry that people put too much stock in the reality of a model, forgetting that it is little more than a theorem that needs to be tested and retested. Contrary to the way some people think, the model is not the endpoint. Rather, it is often only the beginning of the story.

 

Ironically, it is not the modelers who are to blame for the community's overblown faith in these pretty pictures. They are the first people to tell you that their models are only as good as the data they plugged into their algorithms. No, it is inevitably people higher up the pharmaceutical food chain who ascribe "eureka" status to models as a way to explain why their IT budgets are so high. Or media folk who are looking for stories that will write themselves, taking advantage of the picture/thousand words paradigm.

 

What events like the biennial Critical Assessment of Techniques for Protein Structure Prediction (CASP) indicate is that we're still not quite there. Oh, we're getting better at predicting the folding and behavior of proteins and small molecules in isolation, but the data just isn't there for us to get a good handle on what happens in a complex, a cell, or an organism. And the cross-referencing of all the PubMeds and PubChems of the world won't likely be enough, because I don't think the problem is in how we analyze the data but rather the paucity of the data.

 

The Herculean effort required to generate and collate the data required to make half-way decent models will make the Human Genome Project look like a Betty Crocker bakeoff. I just don't know that we have the human and financial resources to pull it off.

 

Do I think we should stop trying? No, of course not. But what I do advocate is that we try to be more reasonable in the questions that we ask.

 

One of the lasting legacies of the Human Genome Project was that it failed to pay the immediate dividends that everyone promised. We quickly realized—or some of us did—that the sequence information, in the absence of the corresponding biological and biochemical information, was largely useless.

 

When you begin to contemplate the so-called human proteome or metabolome, you begin to realize how straightforward the genome was. For all its myriad nucleotide pairs, the human genome is a relatively static entity that is largely identical from cell to cell, person to person and moment to moment. Whereas my genome will be largely the same on Monday as it is today, components of my proteome and metabolome may change significantly if I go to that party and have a few drinks (response to alcohol) or drive rather than take transit (response to stress). The permutations are daunting.

 

So please go ahead and use models in your research. Just make sure that you don't lose sight of the fact that while it may look, act and quack like a duck, it is still a model and may actually be a chicken.