Hyderabad: As scientific disciplines go, the field of ‘Deep Learning’ is but an infant. However, it will soon have a disruptive effect on the field of drug design, experts say.
They were speaking at a panel discussion as part of the ongoing international conference on high performance computing in Hyderabad on Wednesday.
Deep learning, a subset of machine learning, functions by imitating the workings of the human brain to process large amounts of data. The artificial neural networks used in these functions have neuron nodes connected together like a web.
This enables deep learning systems to process data in a non-linear manner, unlike traditional programmes which do so in a linear way.
The panel experts on Monday, which included members from both academia and industry, said deep learning has proven instrumental in areas such as modelling protein structures, human genome interpretation for personalised medicine and understanding host-pathogen interactions (HPI).
Devapriya Kumar, associate professor and head of the centre for computational natural sciences and bioinformatics (CCNSB) at IIIT, Hyderabad, said, “For long, if one had to design a drug for a particular purpose, we took existing compounds which might be close to what we want and performed experiments on them. We would eventually arrive at a synthesised compound that has the properties we want. However, using deep learning functions, we can tell the computer what properties we want in a drug and it would arrive at a compound itself.”
Kumar explained that computing models can also help predict how easily these drugs can be synthesised on a large-scale basis. “After all, it isn’t enough for us to have a compound that does what we need it to do. It should be easy to produce for commercial viability,” he added.
Gopalakrishnan Bulusu, principal scientist at TCS Innovation Labs, Hyderabad, said deep learning could be used to understand the human genome better. “We still do not have a clear understanding of how many pathogens infect hosts such as a human body. We need more data and we need to process this data properly as well. Deep learning functions can help in these aspects,” he said.
‘Deep learning’ in drug design is such a new field that around the world, there are only 200 startups working on it. Most pharmaceutical giants have started to work in this field but there too the activity is in the nascent stage. Speaking with Deccan Chronicle, Kumar said, “Three years ago, this industry didn’t exist. In fact, the number of startups, which is 200 today, was only 80 six months ago.”
Kumar admitted there was a long way to go before India is considered a major player in the field. “Most activity in deep learning in India has been in the field of medical diagnostics.” There are a few startups in Hyderabad that do this. Not a lot has been done in drug design. However, I won’t say we have missed the bus or anything. The field is still brand new and there is ample scope for future activity,” he said.
