Dated September 24, 2001
The Telegraph

India's PARAM (PARAM 10000) Supercomputer leaps into the DNA horizon

For a fleeting glimpse of a DNA molecule in action inside a human cell, a peep that lasts a billionth fraction of a second, biochemist Rajendra Joshi sits at a workstation linked to the most powerful computer developed in India.

Scientists at the Centre for Development of Advanced Computing (C-DAC), Pune are preparing to deploy India's PARAM supercomputer for bioinformatics, a research arena that involves analyzing and interpreting genomic data with the help of computers.

PARAM is a parallel processing machine in which computation-intensive problems are distributed across several processors that work in parallel to slash computation time. C-DAC researchers have tailored popular bioinformatics software packages to run on the supercomputer. They are also trying to develop new software that will speed up the routine processing of the torrent of genomic sequence data that is almost doubling every 15 months.

"Our primary task at C-DAC is to get PARAM ready for Indian scientists to use on bioinformatics projects that will require high computing power," says Joshi, Team Coordinator with C-DAC's Bioinformatics Applications Group.

"This kind of a high-end computing resource coupled with bioinformatics software packages is currently not available in India. We're trying to fill in that gap," Joshi told KnowHOW.

As new genomic sequence data floods into gene and protein databanks, a key task that will keep bioinformaticists busy for years to come is comparing different sequences to look for similarities between them. The discovery of similarities among sequences can be used to assign functions to either unidentified genes or to proteins without known functions.

C-DAC researchers L.A. Anbarasu and V. Sunderarajan have developed what is known as a parallel genetic algorithm that they have used for the alignment and comparison of multiple sequences. This technique produces better results than several conventional techniques of simultaneously comparing multiple genetic or protein sequences.

The C-DAC scientists also hope to use PARAM for protein structure optimization a process that can predict the three-dimensional structure of a protein from its sequence. This is one of biology's biggest unsolved problems. While the sequence of animo acids that make up a protein can be identified by studying the genome sequence, it is actually the three-dimensional structure of the protein that is crucially important in determining how the protein behaves in the body. The three-dimensional structure depends on how amino acids align themselves on space. The idea is to find which is the most preferred shape they will take by looking at the forces of interaction between the molecules.

Software packages for molecular dynamics originally developed abroad are also now available on PARAM. C-DAC scientists have tailored two popular packages called AMBER and CHARMM, both developed by independent researchers in US universities, to get them to work on PARAM.

Molecular dynamics can be used for several applications in biology - to understand the structure and dynamics of a DNA sequence, to understand how a drug binds to a cellular component, or how a proteins interact with DNA molecules.

"When performing such complex simulations, we have to mimic conditions so that they come as close to the real environment inside a cell as possible," says Joshi. "Biological molecules are always surrounded by water and ions and other intracellular components and a realistic simulation should take into account such a cellular neighbourhood."

Joshi himself has used a software called AMBER, originally developed at the University of California, San Francisco, to simulate the behavior of different DNA sequences. The simulation has to take into account interactions between several thousands of atoms in the cells.

"A typical simulation of a biological molecule for just about one nanosecond a billionth of a second - in the presence of water molecules and the myriad ions could involve up to 20,000 atoms and on a single processor workstation could take several months," says Joshi. "On the PARAM supercomputer, we would have our simulation in less than a week."

As C-DAC has aready demonstrated, bioinformatics on PARAM may yield spin-off in the understanding of complex diseases. Joshi has used molecular dynamics on PARAM to throw new light on certain sequences that tend to get repeated in the human genome and have long been associated with certain neurodegenerative diseases such as Huntington's disease.

He concentrated on the sequence C-A-G that repeats in a certain section of the genome 6 to 35 times in normal people, but is found multiplied 36 to 121 times in people with Huntington's disease an incurable neurodegenerative disorder. Molecular simulation of this repeat sequence shows that the three dimension structure of the sone of DNA that makes up these repeats is highly flexible. While the neighbouring sequences are relatively rigid, the C-A-G section is flexible. This could be why C-A-G repeats itself so many times.

This simulation of the DNA repeat zone has provided a new insight into the molecular mechanisms that underlie highly repeated sequences linked to this neurodegenerative disorder.

by G.S. Mudur