The Role Of Augmented Intelligence In Fighting Against A Pandemic

Shailesh Manjrekar. July 16, 2020
The Role Of Augmented Intelligence In Fighting Against A Pandemic

Originally published on Forbes Tech Council, May 27, 2020

This article originally published on Forbes Tech Council is republished here by author Shailesh Manjrekar, head of AI and business development, at HPC Storage innovator WekaIO.

In 2015, Bill Gates did a Ted Talk, one that would certainly paint him a visionary today. In his talk, he outlined what the world would need to do to cope with a global pandemic. He pointed out that the world wasn’t ready — there was still much that needed to be done to protect humanity and economies.

The ability to scale up diagnostics, drugs and vaccines very rapidly is paramount. The technologies exist to do this well if the right investments are made. These investments could mitigate the potential $3 trillion hit to global wealth and millions of deaths — a figure put forward by the World Bank back in 2008.

Unfortunately, 2020 has brought the pandemic that Bill Gates discussed, and the majority of countries around the world were caught off guard. However, technological advancements have certainly happened, and we are in a better position to fight back than, say, when the Spanish flu hit in 1918. Artificial intelligence, machine learning and deep learning in particular, will play a vital role going forward.

Advancements in artificial intelligence and deep neural networks are helping medical professionals in their increasingly busy roles. What has come to be known as “augmented intelligence” — where AI is used under human supervision — is being deployed for effective COVID-19 diagnostics. The effectiveness of such diagnostics is very much dependent on the datasets we have access to — hence, global collaboration is paramount to accumulate diverse, rich and demographic-specific datasets. Equally important is explainable AI, which is the ability to explain an experiment by using data versioning, audit trails, lineage and annotations. In my role at Wekaio, I get to talk to several of the healthcare professionals and data engineers who manage these datasets.

Next-Generation Genome Sequencing (NGS)

NGS has progressed where the cost of whole-genome processing has come down below $1,000 and made it close to real-time, leveraging the Genome Analysis Toolkit (GATK) graphic processing unit (GPU) acceleration and Field Programmable Gate Array (FPGA) acceleration as adopted by startups Parabricks and Edico Genome, respectively.

Additionally, 100 GB-per-second networking and flash-based software-defined storage can handle small and big FASTQ and SAM/BAM files, effectively, during secondary and tertiary analysis. This enables access to the most important asset, the dataset, and speeds up the process of analyzing a whole genome in one hour.

When it comes to COVID-19 and genome samples from patients, sequencing has discovered eight strains and 30,000 base pairs, confirming that the mutations are not much different from the original strain. The results gleaned from these analyses are uploaded to to enable open-source, global collaboration and to improve outbreak response.

The ability to scale this processing is crucial, as predicted in a PLOS Biology study conducted in 2015.

Augmented Intelligence In Clinical Diagnostics

AI is making a difference in the clinical diagnostics of X-rays, MRIs and CT scans. Radiologists tend to look at multiple artifacts to determine an anomaly and predict the probability of a patient developing COVID-19. Deep neural networks (DNN) can be trained to detect specific anomalies as long as there’s reliable data. These datasets need to be annotated and labeled, which is where technologies such as transfer learning can assist in training the models on one kind of dataset and then fine-tuning using actual datasets.

Collaboration is a challenge today, due to stringent privacy and ethical requirements pertaining to sharing patient data. This is where federated learning helps: Instead of sharing data, models are trained locally and tuned globally. Additionally, software-defined storage solutions with in-flight and at-rest encryption help eliminate security and compliance concerns.

Several startups are fueling this innovation, with the likes of Infervision and Vuno leading the charge. There’s also India-based, which has been processing up to 5,000 scans a week, and DarwinAI has developed an open-source convolutional neural network called COVID-Net that uses explainable AI to annotate X-rays to augment prediction summaries so radiologists can better understand the predictions, providing clinicians with information and insight into the disease.

Augmented Intelligence For Cryo-EM Data Selection And Classification For Drug Discovery

A high-resolution imaging tool called cryo-electron microscopy (cryo-EM) flash-freezes molecules in liquid nitrogen and bombards them with electrons to capture images with a specialized camera. This leverages molecular modeling to solve the structure of complexes in days, including the 3D structure of spike proteins on SARS-COV-2 and how they interact with human proteins.

Convolutional neural networks (CNN) are commonly used with cryo-EM to help eliminate false positives. However, determining the structure of a protein in this process requires millions of samples. This is where techniques such as positive-unlabeled learning can use a small number of example protein projections to train a CNN to detect proteins of any size or shape.

AlphaFold, an open-source deep-learning library from Google-owned Deepmind, uses DNNs to predict the protein structure based on their genome. Computational chemistry startups like Benevolent AI use AI to look for existing drugs and their impact on COVID-19.

Covid-19 HPC Consortium

The White House and the U.S. Department of Energy have created the COVID-19 High-Performance Computing Consortium with support from multiple technology companies and universities. The goal is to bring 330 PFLOPS of compute power and high-performance storage to focus on COVID-19 research. And AWS just announced that it will make COVID-19 datasets freely available in the AWS data lake.

These systems allow researchers to run large numbers of experiments in epidemiology, bioinformatics and molecular modeling. AI is also used for providing transparency and containment using machine learning, robotics and autonomous machines.

In The End

Augmented intelligence is the foundation upon which the future rests. It is on track to change the way the world manages a pandemic and the tools it can use to detect and cure potential outbreaks in the future. As governments and the private sector continue to collaborate on these solutions, their investment will shape a world that’s capable of protecting its people and economies when a crisis hits.

Additional resources:

Related Resources

Case Studies
Case Studies

Oklahoma Medical Research Foundation (OMRF) Case Study


Datasheet - Hitachi Content Software for File

Solution Brief
Solution Brief

Hitachi Content Software for File: High Performance Storage for AI & ML