An AI platform that predicts future variants of SARS-CoV-2.

DMT is building an AI-driven platform that can predict where and when in the genome the SARS-CoV-2 viral strains will mutate, enabling the efficient development of vaccines and therapeutics.

Dineo Lioma, CEO of Deep Medical Therapeutics. 

Africa is currently a COVID-19 vaccine and genomic data “dark spot”, accounting for only 1.3% of the global Sars-Cov-2 genomic data. There is also a lack of understanding when it comes to this existing genomic data, especially concerning the emergence of new, deadlier strains, such as the (501Y.V2) South African variant which threatens to stall progress on global vaccination programs. This genomic data is important (1) For Health ministries to understand the viral strains circulating within a community (2) To enable pharmaceutical companies to develop effective COVID-19 drugs and vaccines. 

Africa is at high risk of being left behind when it comes to vaccinating its population. We need 1.5 billion vaccine doses to immunize 60% of our inhabitants. The major concern is the efficacy of the vaccine procured for use in Africa. Since the emergence of different variants such as the South African (501Y.V2):

• 1.5 million doses of the Oxford-AstraZeneca vaccine were halted due to only a 10% efficacy rate. 

• Moderna found a sixfold reduction in the antibodies’ effectiveness  

• Pfizer stated its vaccine may be 2/3rds less effective. 

We need tools that leverage genomic data and AI to predict future COVID-19 variants which will inform future drugs and vaccines design.

To better understand how our solution will serve our end-users needs, we have held interviews with two large pharmaceutical companies, the African Union - NEPAD, South African National Biodiversity Institute, and the National Health Laboratory Services of South Africa. We are also an affiliate member of the Avoca Quality Consortium which is a global leader in the implementation of clinical trial quality. Below we outlined the needs our solution meets for two of our key stakeholders.

Governments need:

• Data-informed decision-making when it comes to disease outbreaks.
• Data to inform drug and vaccine procurement, ultimately saving their countries money through minimizing wastage (e.g. procuring vaccines that have high efficacy based on the strains circulating locally).
• Support to reduce the transmission and spread of disease by providing data for effective intervention measures.

Pharmaceutical companies need:

• Access to genomic data insights from Africa.
• Support with the clinical trial site and candidate selection.
• To reduce up to 50% on the cost of development of drug or vaccine.
• To save between 6-11 years on time to market for a drug or vaccine development.
•  To assess vaccine efficacy and have data that will enable them to develop more effective vaccines.

All research and findings discovered from our predictive COVID-19 mutations platform will be made publicly available and shared with the African Union to support their strategy to fight COVID-19 and their vaccine procurement plans. This information will also support pharmaceutical companies in developing higher efficacy vaccines; directly saving lives, time, and money. DMT will also publish a white paper on the scientific findings of our research and AI model contributing to the research community. 

Africa has the fastest-growing population and bears 24% of the global disease burden. DMT aims to empower Africans with access to tailored healthcare solutions at an accessible cost. 

Our software has the potential to support fully eradicating various diseases in Africa and around the world. The World Health Organization outlines this as being a key element in achieving the sustainable development goal, SDG 3, on health and wellness.

Africa is at high risk of being left behind when it comes to vaccinating its population, especially given its social, political, and economic vulnerabilities. Our solution will help Governments safeguard their citizens from COVID-19 by proactively identifying new viral strains that decrease vaccine efficacy. It will also help Pharma in developing effective vaccines, saving time and money in research and product development costs.

Our Activities:

• Working with the South African National Health Service.
• Interacting with African researchers currently sequencing COVID-19.
• Use public research and data to build a predictive AI model.

Our Outputs:

• Predictive AI models providing where and when a COVID-19 mutation of significance will occur across the globe.
• Scalable intelligent software that will expand to other disease areas e.g. HIV, TB, Malaria.

Our Outcomes:

• Data-informed decision-making for Governments on their COVID-19 response strategy and vaccine procurement.
• Reduction in money lost due to procurement of vaccines that are ineffective for Africa.
• Data to support effective vaccine development. 
• Lives, time, and money saved.

Evidence to support our claims includes the history of the Rotavirus Vaccine which was developed using viral strains found only in Europe and North America but not in Africa and was less effective in our continent.

Over the next year we intend to:

1. Refine our AI model with more SARS-CoV-2 whole-genome sequence data to improve the accuracy of its predictions of variants from Africa.

In the next 1-3 years, we intend to scale our impact by:

2. Expanding the AI platform to other parts of the world so that it can be used globally. 

3. Sharing insights generated to date with the pharmaceutical industry to help with R&D of vaccines. Insights will also be shared with Governments to support the surveillance and planning.

4. Increasing the data access by increasing the number of data contributors on the platform.

5. Scale the platform to other disease areas that are prone in Africa such as TB (502/100,000 people), HIV (5.49 million people), Cancer (49% increase by 2030 compared to 2012), and Hypertension (6.3 million). Insights discovered will inform local, national, and international actions needed to prevent, prepare for and respond to threats related to diseases in Africa.

Many researchers do not fully understand how the virus is evolving in human hosts & pharmaceutical companies are not able to develop high efficacy vaccines that will still be robust against future variants. We are unlocking the full value of genomic data by using AI to better understand the evolutionary behaviour of COVID-19 and predict future mutations of significance that would alter vaccine efficacy. This will support governments in making informed strategic decisions and help pharma develop effective vaccines when the virus mutates as it replicates.

Our impact is tied to the implementation of these outputs which also requires collaboration with the government & pharma stakeholders. We are building partnerships with the African Union, The South African National Health Services and pharmaceutical companies to ensure any insights are used by these stakeholders, ensuring impact. Below we list some key impact indicators that are within our control.

• Built retrospective model accurate to >80 % (current platform predicts the South African (501Y.V2) variant with >50% accuracy) 
• Built AI perspective model accurate to 80% 
• Software’s ability to provide actionable future insights that are shared with the SA government
• Assessment of vaccine efficacy with mutant strains and shared with pharma companies

Genomic data access: Limited genomic and epidemiology data to feed into the DMT software. Africa only contributes 1.3% of the global COVID-19 genomic data, which have been uploaded onto public databases like GISAID. However, we plan to partner with other organizations such as the National Health Laboratory Services (NHLS), KRISP, and academic institutions around Africa conducting sequencing to get access to further data which has not been uploaded on GISAID. The more data we have the more accurate the predictions the model makes.

Regulatory: Lack of regulation around the acquisition, use, and sharing of genomic data in Africa. We will use already publicly available data. In the future, we will ensure broad consent and permission for secondary use of genomic data is attained. We are also appointing a health data and ethics advisor.

Financial: Costs of infrastructure and the development of AI-driven platforms are high. We started by leveraging free software for our proof of concept. We have an engineer and virologist who are working for free to help with the initial stages of development until funds have been raised. We are fostering connections with the Wellcome Trust, B&M Gates Foundation to try to secure grant funding. 

The Avoca Quality Consortium- an organization comprised of over 120 pharma, biotech, CRO, and clinical service providers with the shared objective of elevating the clinical trial quality and bringing key stakeholders in the clinical trials process into greater alignment.

Barrier: Lack of Financial. To date, we have used our own personal savings and leveraged pro-bono work. However, this is no longer efficient as we need a focused, full-time team that can deliver the results at the pace we require. By receiving a monetary prize from the Trinity challenge, it will allow us to expand our team and invest in the infrastructure and human resources required to develop and scale this technology.  We would also like to connect to and work with some of the partner organizations and companies within this powerful network which we believe will accelerate our growth trajectory. 

Global Virome Project: Which is an organization focused on the discovery of zoonotic diseases such as COVID-19 and stopping future pandemics, could assist us in shaping and deploying our solution. 

Discovery Health: Is one of the leading Health Insurance companies both in SA and globally. It has been at the forefront of using data and mobile technology to help guide the country on COVID-19 interventions with its COVID-19 Alert SA app. We could partner with Discovery to share insights on the COVID-19 viral strains circulating within a community.

Microsoft and Google: Our platform will require substantial computing power and storage once we scale; as market leaders in this field, these organizations can offer guidance and support on which solutions we should use. These organizations could also offer the expertise we require in machine learning.

University of Cambridge: To help guide our project, we would like to engage with the Prabakaran Group, the Micklem Lab, and the Pathogen Dynamics group which focus on applying machine learning to big genomic data, genetics, and the epidemiology of infectious diseases. 

Bill and Melinda Gates: For potential funding and connections with its member organizations and grantees that would support the development and deployment of our solution.