Technology has been a major catalyst for revolutionising the healthcare industry in the past few years. From diagnosis, to patient care, up to the formulation of vaccines, technology granted medical practitioners new ways of saving people’s lives, while also enabling patients to become partners in their well-being.

With new developments in the field of genomics, experts are now looking at transforming the field of healthcare from reactive to predictive— an objective that is fast becoming a reality through technology.

During a keynote entitled “Combining High Performance Computing, Genomics, and AI to Enable Precision Medicine,” as part of the latest Healthcare Frontiers online conference organised by Jicara Media, Ananda Bhattacharjee, Head of Business Development and Solution Architect for HPC and AI at Lenovo, illustrated how precision medicine can be achieved through the combination of HPC, genomics, and AI.

“We want to get to a point where we can treat you, to measure the individual susceptibility of your diseases, or protect you, your environment, (and) your response to a specific treatment. We want to know before these things (illnesses) happen,” Bhattacharjee said.

“We want to take into account your genetic background (and) your lifestyle, so that we find the right drug and dosage for you, and tailor healthcare to maximise the benefit and minimise the harm. We want to move away from treating the symptoms after it has happened to knowing before it has happened,” he added.

The journey of precision medicine

According to Bhattacharjee, precision medicine was born around the time that scientists were trying to sequence the first genome in 1990.

“It took around 13 years to sequence the first human genome. And it took around $3 billion. But quickly, once it has been done, the scientists (understood) that sequencing a single genome is (not enough to) decipher the secrets of your health. To find the secrets to longer life, we clearly realised that we needed to sequence many of us. We needed to sequence a multiple of us to get more insights. But together it is a challenge— it is mapping human variation, the differences we can see. Genome to genome from individual to individual is a variation, (and) that variation results in susceptibility to diseases. It can be as simple as a height,” he explained.

Today, genome sequencing no longer takes 13 years to accomplish, and the costs have also significantly gone down.

“We see (genome) sequences everywhere. (Genome) sequencing costs (have) come down to less than US$1,000. I’m talking only about the sequencing part, not the analytics part (yet). We see genomics in the lab, in the virology and agriculture fields. Plant genomics (is used when) we need a new variety (or) breed of plants. The balance has shifted now from the next generation sequencing side, to more on the analytic side, with the cost per genome coming down drastically from the sequencing side. So, it’s now the analytic side, which is more and more important now, to bridge this gap of sequencing up here (in) our human level population,” Bhattacharjee noted.

Precision medicine has also been instrumental for scientists during the ongoing pandemic, not only for the development of COVID-19 vaccines, but also for studying the behaviour of the virus.

“While there are many unknowns while we talk about COVID, (we can) say (that) researchers are tackling it from multiple places, like tracking the virus origin, or it may be vaccination design. One of the first steps to scientific insight is -omics analytics. In fact, we see that as necessary for any kind of research which is happening in the area of COVID-19,” he said.

The role of HPC, genomics, and AI

The process of genome sequencing, according to Bhattacharjee, starts with a biological sample such as blood, saliva, or tissue, which is then loaded onto a sequencer that converts the sample into digital information.

“Think of it as a puzzle which doesn’t give you a (complete) picture. The information is there, but you can’t make it out until the picture is there. That is what we call genome analytics, where HPC, genomics, and AI play a very big role. That’s where we require a supercomputer on an HPC, which can do this with a good analysis software,” he pointed out.

“The whole workflow starts with sequencers, taking the data out from the samples, (and) giving you raw data (that) needs to be analysed, to find out the properties of the genome or the characteristics of the genome. In the bioinformatics world, this is called variant analytics, where we compare the genome with a standard genome, and find the variants. From the variants we find the characteristics— the phenotype and the genotype,” he added.

While genome sequencing has contributed huge strides in healthcare, the process of genome analytics, which follows genome sequencing, is beset with a major setback.

“The bottleneck currently, which we are seeing, is the amount of time it takes to analyse a single genome. If you’re talking 60 to 150 hours for analysing a single genome, it’s like one week, (and) this cannot be done if you are talking of a population-level genomics,” Bhattacharjee said.

How then to proceed towards a precision healthcare paradigm from a traditional one? Bhattacharjee enumerated five ways:

• Make genome processing fast.
• Increase throughput of genomics analytics.
• Make it affordable.
• Make it easy to use.
• Make it secure.

As such, Lenovo has introduced its GOAST solution. GOAST, which stands for Genomics Optimisation And Scalability Tool, can analyse up to 27 genome samples per day, according to Bhattacharjee.

Lenovo GOAST comes in two forms— the Lenovo GOAST Base, and the Lenovo GOAST Plus. For the Lenovo GOAST Base, it takes about 3.3 hours to analyse one genome sample, therefore capable of analysing 7.3 samples per day, or 2,700 per year.

For Lenovo GOAST Plus, it takes about 53 minutes to analyse one genome sample, yielding up to 27 processed samples a day, or 9,700 per year.

“We spent a lot of time in our lab, understanding the characteristics of the genomics apps and benchmarking it against the computing elements. So it’s an interplay between the hardware and the software, so that the researchers don’t have to spend time, and the bioinformatics guys don’t have to spend time on paralysing applications. They can start on the work on genome analysis from day one,” Bhattacharjee emphasised.

“If you don’t do genome processing, you’re not talking about precision medicine, because you are not tailoring a particular individual’s genome. So for that, what we do is we make genome processing faster. We talked about 167x, where you can take off population-level genomics. You can analyse hundreds and thousands of genomes per year. We want to make it affordable, we call it ‘GPU level speeds at CPU level costs.’ We make it easy with pre-configured and pre-installed systems, which will help you to plug and play and run, and just talk about the computer science part of it,” he added.

Since genomic analysis involves a lot of confidential patient data, Lenovo has also ensured that its GOAST solution addresses security concerns.

“We make it secure. You run it on-premises in your organisation, so that you don’t have to deal with security issues when it comes to human data and putting it onto the cloud,” Bhattacharjee said.

Going forward, a lot more has to be done to mainstream the practice of precision medicine, Bhattacharjee stressed.

“We need to have variables, where it can capture data which talks to these ecosystems. We need to improve the natural language processing, so that we can mine the electronic health record. We need to do a lot of work to wrangle the data, to curate the data mine— it’s an ongoing process, and we as a society need to work much in this area,” he remarked.

“I can’t think of a better area where technology can help the medical field and make a real change in the human population. It’s really an area where technology can make a huge change for the generations to come,” Bhattacharjee concluded.