Blast from the Past: Computing the Hard Way

When biochemists Marv Hackert, Jon Robertus and Steve Ernst began using the Vector General computer graphics system (pictured here) in 1978 to assist with modeling protein structures, it was at best a very limited and very expensive tool.

For the roughly $250,000 which the university spent on the system, researchers got black and white images plotted in straight lines (no curves), and they still had to do most of the work themselves, entering the thousands of coordinates into the computer in order for it to do its work. The computer was less powerful, says Robertus, than a $400 PC is today.

“Our attitude was: that’s nice, but it’s slow,” says Hackert. The computer was almost an afterthought to the primary process of modeling new protein structures in a Richards Box—also known as “Fred’s Folly”—which used a light bulb, a mirror, and a series of stacked transparencies to transfer the information from X-ray crystallography into a ball-and-spoke model of a protein molecule.

The process took months, and feeding the coordinates into the computer could add another month. The computer representation was helpful, says Robertus, but “not essential.” Not until the 1990s did the software and hardware become sophisticated and powerful enough to really revolutionize the field.

“Now we can feed an electron density map into a computer and have a model in less than an hour,” says Hackert. “In the early days, it was like the California Gold Rush. You’d be in the stream and you’d find a nugget and hold it up to the light and appreciate it. This was when we were just starting to solve these structures, and every new structure was an event. Now, when we have tens of thousands of structures, it’s like blasting a mountainside, and washing away the dirt with high-pressure hose. It’s all automated. You never get the joy of holding up the nugget.”

“We have to be careful not to be too curmudgeonly,” says Robertus. “Nobody wants to go back, but it is true, I think, that it’s now so easy that people tend to be interested only in the end product.  Back then you had to know what you were doing when you were doing it by hand. And in the early days of computer modeling you also, often, had to write your own software. You developed a fuller understanding of the underlying science.”