Genomic Sequencing Isn’t Difficult Anymore: Sinisa Nikolic, Lenovo ISG

In conversation with Sinisa Nikolic, Director and Segment Leader, HPC & AI, AP, Lenovo ISG, BW Businessworld explored the world of High Performance Computing (HPC) in conjunction with healthcare and genomic sequencing. Read on for excerpts from the interview.

Edited Excerpts:

What will be the role of AI and ML in moving towards preventative healthcare? 

AI has been used in healthcare for 5-7 years now. But it wasn’t exactly focused towards preventative medicine. For example, it has been applied in oncology for treatment paths. We have seen a complete and utter ballooning of this segment of AI application in healthcare market. These are on-point, just-in-time treatment machine learning applications.

From the predictive analysis perspective of AI and what’s really cool about it is – the predictive ability based on the algorithms that we use. Today, we have the ability to be able to predict healthcare-related issues.

Lot of governments are doing larger population-based machine learning (population-based genomics). To start drawing inference from that level of data whether it's for a specific sex, cultural group, location, socioeconomic, etc., you actually need AI tools.

With AI-based, more cost-effective HPC infrastructures, governments and organisations are spending a lot lesser on per day basis in comparison to spending millions on HPC 5-10 years ago.

Why has it been so hard to go about genome sequencing? What are the challenges involved for technology in this? 

Genomic sequencing isn’t difficult anymore. A sequencing tool, today, can take 100-150 hours to sequence someone’s DNA. So, I would say that it’s not that it’s hard – but it's actually expensive, still. It's speeding up and getting better. And as more of Next-Generation Sequencing (NGS) systems come out, these things halve the sequencing time constantly.

Do we finally have the technology that can bear us fruits in the speed that humanity needs? 

Yes, because things are happening very quickly. At Lenovo ISG, we've just released the next version of GOAST and it continues to reduce time to these insights. The process used to take days, 150-200+ hours. Now, it’s been brought down to minutes. We've seen 167x increase in speed.

Depending on what data set we are processing, we've seen 150-180 hours come down to 48 minutes and even seen it drop closer to 12 minutes. Now, what's really interesting by that is that we don't use GPUs, which are very costly. And whilst I talk about systems getting vastly powerful, they're also getting cheaper.

Has HPC, AI and cloud computing received a greater support since the onset of the pandemic? Have you observed a greater mobilisation and support for HPC? 

That's actually a brilliant question from the perspective of everyone. We had so many clients globally that had to pivot incredibly fast to support so many remote workers. Networks were stressed, the internet was stressed, everything was stressed. What we saw from a HPC perspective was growth.

Also, we saw government­-to-government collaborations for genetic sequencing on makeup of the virus. We went through sequencing, the data was divided, a little bit like city at home. But everyone worked on it. We saw sustained focus within HPC and AI.

Could you tell us about your efforts to advance cancer research in India?   

With our technology, CSIR-IGIB research team were able to implement GOAST. So, we were able to start sequencing specific cancer genomes and digging deeper into the genetic roots. We deployed a fairly large system – the largest GOAST installation in India to date, across 28 nodes.

With our work with CSIR-IGIB, we took the process of genome sequencing from 60 hours down to nine hours. This means that the researchers are able to push more scientific work through, they're able to increase the complexity of the questions they ask, because they have technology in front of them that is able to handle those types of things.

In the context of genomics data, do we have to be worried about cyberthreats?

When you look at healthcare services and research organisations, they store an incredible amount of patient data genomics data. The data is confidential and valuable. But what can a cybercriminal do with this data? Today, not a lot. But as you move forward, this DNA is our fingerprint, it's our life. It sort of like a PIN number, something that we can’t change. So, in near future, if we could have genetically stored passwords or banking encoded to our DNA. With sequencing getting faster and faster, stolen genomic data could be a big threat.

Approximately two billion human genomes will be sequenced by 2025. So, this data will become interesting for cybercriminals.

Also Read: Attack Surface Visibility, The Key: Vijendra Katiyar, Trend Micro