MATHUKUMALLI VIDYASAGAR: We work side by side with the

biologists and help them to interpret and get knowledge of

all the experimental data that they're generating. That's

what computational biology is about. My current research is

called computational cancer biology. What we are talking

about is once you get cancer, what should be the therapy?

I spent about a year or two talking to cancer doctors and

say what is it that you see as the biggest challenges? So it

took me a while to figure out what is the intersection, the

question that bothered them and the question that I can answer

and finally we zeroed in on this personalized cancer therapy for

specific forms of cancer of the uterus. Can you predict which

patients will respond well and which patients do not respond

well to specific therapy? So they had certain guidelines that

when the patient's tumor was more than two centimeters in

diameter, in addition to taking out the uterus and all the

associated parts; they also removed the so called lymph

nodes for fear that the cancer had already spread there. And

then when they did the analysis after the surgery, over a very

long period they discovered that 78% of the surgeries were

unnecessary. So they had perfect knowledge that most of the

surgeries were unnecessary but no way of predicting beforehand

which were unnecessary. So this was the challenge they posed to

us. Can you find some indicators? So this took about

six to eight months of algorithm development, new computational

methods and then we had reasonably well working

predictive procedure. Then we collected about 28 new tumors.

We applied our predictive methodology on those and out of

these 28 people, only nine of them really required surgery of

the lymph node, and we were able to spot eight out of

those nine. So they're really happy with that.

There's still a lot of people out there who think

that the way to solve problems of cancer is trial

and error. So to change the mindset and say you don't

have to rely on trial and error, you don't have to rely on

serendipity, you can actually undertake systematic analysis of

large amounts of data to come out with plausible hypothesis,

only a few people accept this now. So I'm hoping that through

our work we'll come to a situation where this approach

essentially becomes natural. People should say why would you

want to do trial and error? This is the way to go.

♪ MUSIC ♪