Profiles in Computer Science Courage Part II: Advice on Taking the Plunge
Words of Advice from the Scientists Featured in Profiles in Computer Science Courage
Find Your Passion
“Not every computer scientist will fall in love with the field like I did,” says Ron Shamir. “And that’s an essential part of doing research: to be fascinated, excited and enthusiastic about what you do.”
It helps, Bruce Donald says, to find a great lab where people are doing computational biology that excites you. “You must develop the ability to admire the work and decide whether you’d like to do work like that.”
When he was a master’s student at Stanford, Michael Black told a professor he wanted to do his PhD in cognitive science and study human perception. The professor told him, “If you were my brother, I’d tell you to get a computer science PhD because you’ll make more money.” Black took that advice, but managed to find his way back to cognitive science through the study of computer vision. The advice he gives his students is different: “Follow your heart.”
Haussler agrees: “You’re limited only by your passion and commitment.”
Learn Some Biology
Certainly a computer scientist who wants to work in biomedicine must learn some biology. The question is, how much? According to Shamir, “Initially, a computer scientist can pick up what he or she needs to know about a biological problem by reading chapters in one good book. To get more seriously into the field, one has to attend conferences and follow the recent literature.”
Daphne Koller suggests that instead of taking introductory biology classes, which can be descriptive rather than quantitative, computer-science graduate students should start by reading a more advanced textbook and some more computationally oriented papers in the good journals. “Get a sense for the kinds of work people are doing,” she says. “Find a problem that interests you and then find the background courses and reading you need.”
When Gene Myers made the leap 30 years ago, “I was lucky enough not to have to know a darn thing,” he says. “I would have a conversation and do the best I could.” But now, because the level of sophistication in computational biology is increasing, he thinks more is needed. “Take some biology courses or go study with somebody in the field. I think at this point that’s a requirement.”
Donald agrees. In the early days, he says, people felt you didn’t need to know a lot of biology and biochemistry to pick a deep problem and work on it. “I’m not sure that’s true anymore,” he says. “I’m not sure it’s good enough to learn a little.”
Find Great Collaborators
Computer scientists agree that working in biomedicine depends on personal connections with biologists you can trust. Virtually all of those profiled here say they had great collaborators early on. “You need people you can ask, ‘am I doing the right thing?’” Leonidas Guibas says.
Today, because there are more biologists with a quantitative background, it’s easier to find people who both understand what computer scientists can offer and speak the same language, Myers says.
But beware collaborators who have a naïve view of the computer scientist’s skill set, says Black. “They might see the computer scientist as the programmer who comes in and writes some code,” Black says. “Collaborators should understand that a computer science collaborator brings ideas and ways of looking at the problem and understanding the data and maybe whole new ways of thinking about what the biological system is doing.” Computer scientists also shouldn’t make the mistake of seeing biologists as a source of data, Black notes. “Collaborations require people to appreciate each other.”
Experiment with Experiments
All of the students in David Haussler’s lab have the opportunity to work in his wet lab. He doesn’t expect that the computer science-oriented students will remain there, but many do a stint out of curiosity and to broaden themselves.
“It’s important to learn and understand the other person’s language, concepts and worldview,” Haussler says. In the end, some might find they are adept at both the pipette and the keyboard. “They can lead a complete and rich double life,” Haussler says. “But not everyone has to do that to be successful in this field because we can do work in teams with people who complement each other.”
Ask Lots of Questions:
“Ask the right questions and don’t assume you know the biology,” Paul Groth says.
A little biological knowledge is a dangerous thing. “You may miss something important when helping [biologists] design new systems or designing new computational approaches to what they’re doing,” he says.
“If you only want to prove theorems, you will not get very far in biology,” Shamir says. “You have to compromise: if you can’t provide an elegant formal solution, you should be willing to sometimes work with heuristics and algorithms for which you aren’t able to prove much. And you have to work with real data and interact with biomedical experts who think differently and have different goals. This requires adaptation.”
“The best way to learn is to teach,” Shamir says. “Teaching in a different discipline is hard, but it is also very rewarding.” Preparing course materials can be a bigger commitment than just writing papers, Shamir says. That was especially true in the 1990s when he started creating bioinformatics courses and there were virtually no textbooks. “I had to create the course out of primary journal papers. It was very hard work – but it taught me a lot and was fun.” And lecturing is a learning experience too. “You interact with young minds, force yourself to organize your knowledge systematically, and through the process come up with new research questions.”
Be Prepared to Slow Down
Some computer scientists find the pace of experimental science frustrating, Black says. Particularly in cognitive science, the area in which Black works, it takes time to train animals, perform required surgery, deal with governmental regulations, and obtain experimental observations. Human studies can be even more frustrating. People leave the study, patients die, “many things are out of your control,” Black says. “So I’ve had some computer science students back away from
the biology to stick with computer science.”
Mind the Gap
When you work in interdisciplinary science, Guibas says, you have to decide what community you want to be part of. “There’s a danger of falling in the gap between fields. Your work might be too computational for biological publi- cations or too biologically specific for computer science publications.” Computer science done for a biologist might end up in the fine print at the end of a biology paper, Guibas warns.
Be Both Bold and Careful
In computer science, Donald says, people are excited about creativity, spontaneity, innovation and boldness. “Computer science has an element of surprise,” Donald says. “You’re trying to make the computer do something that it couldn’t obviously do before, such as redesign an enzyme to have a novel function.”
Experimental scientists have a different set of values built around being careful and controlled, Donald says. They dot their i’s and cross their t’s. Being bold, as computer scientists are wont to do, might seem risky to them, Donald says.
But really, the kind of care that is necessary to doing biology can be useful in computer science, Donald says. And at the same time, “the bold- ness and creativity of computer science in the hacker generation can be really exciting for trying new approaches in biomedical research.” Computational biologists can do both: be bold and careful at the same time.