New Solutions to Climate Change, Lab-Grown Meat, and Covid-19 with Machine Learning

For years, YCombinator (YC) has published a Request for Startups (RFS), calling on startups that are working on some of today’s biggest problems to apply to the selective startup accelerator; problems like climate change, mass meat manufacturing and Covid-19. In a series of blog posts, we are going through each one of the items on YC’s RFS to discuss the role of machine learning in bringing society closer to closing the gap between these problems and their solutions.

Last week we covered the first three items on the list: AI, Bio, and Brick and Mortar 2.0. Today’s focus is on carbon removal technology, cellular agriculture and clean meat, and responses to Covid-19. 

Carbon Removal Technologies

Scientists have found that the leading cause of earth’s rising temperature is the increase in greenhouse gases in the atmosphere, most notably carbon dioxide (CO 2) and methane (1). One potential method of removing CO 2 from the atmosphere is through direct air capture (DAC). DAC is when air is pulled into a machine and CO 2 is filtered out. In some cases, upwards of 80% of CO 2 is removed. After the CO 2 is separated it is stored underground or used for products like fuels (2). 

The following are a few ways machine learning is being used to help advance innovations in direct air capture:

Photo by Maxim Tolchinskiy on Unsplash

1. Development of novel component materials: DAC devices pull in air and filter out carbon with a CO2 sorbent, which is similar in concept to a sponge that absorbs gases. Researchers Andrew S. Ross and Evan D. Sherwin from Harvard University and Carnegie Mellon respectively wrote on the topic of material discovery for sorbent design, “While CO2 sorbents are improving significantly, issues still remain with efficiency and degradation over time, offering potential (though still speculative) opportunities for machine learning. Machine learning could be used to accelerate materials discovery and process engineering workflows to maximize sorbent reusability and CO2 uptake while minimizing the energy required for CO2 release.” (5)  

   Combining data on molecular structures and their behavior or properties, along with data regarding the laws of physics, materials scientists can create models that predict, given a certain molecule, what is the likely behavior or property of that molecule. These insights are valuable when determining what chemical ingredients to use for a certain type of material given the desired specificities of that material, in this case, the ability to absorb C O2 from the air. This can also be used in the case of lab-grown meat, where various combinations of chemicals makeup the nutrients used to grow cells (see more on cellular agriculture below.) Find more information on materials discovery here. 

2. Characterizing geologic resource availability: machine learning can be used to determine where the best geographical locations are for CO 2 that has been removed from the atmosphere. Several characteristics go into determining if a site is appropriate including whether it is able to accept large volumes of C O2 and store the C O2 for a long period of time (3).
3. Monitoring underground C O2 in sequestration facilities: Once CO 2 is injected into a storage site, machine learning can be used to monitor and maintain that site. For example, computer vision can be used to detect leaks in the C O2 containment or other methods of machine learning can use sensor data to help predict if there is still room left at the site for more injections of C O2. 

Cellular Agriculture and Clean Meat

Second on the list is cellular agriculture and clean meat, also known as lab-grown meat. Lab-grown meat is the process by which scientists harmlessly extract cells from animals and then mix those cells with additives to grow meat-like products. Among the environmental and ethical concerns around eating meat, many believe that our current production methods simply won’t meet the growing demand. In fact, the world is expected to eat more meat in 2021 than ever before (4). An alternative to traditional meat manufacturing is meat grown in a lab.

The following are a few ways machine learning is being used to help advance innovations in lab-grown meat:

1. Analyzing large amounts of text: Cellular agriculture is still in the early stages of discovery. While lab-grown meat has been approved by regulators and sold in Singapore, it has yet to hit the commercial market in the U.S. Developing new technologies like lab-grown meat requires extensive research and development work. Artificial intelligence is being used by some companies in the scientific community to expedite the research process. Iris.AI is a company that has created an AI research assistant using natural language processing that helps manage scientific knowledge. 

   If given a certain paper, topic or problem statement, the language model (trained on large amounts of scientific text) categorizes the mass of results and even gives a score of how relevant an identified paper is. This can be used across industries, but in particular for cellular agriculture where there are still numerous technological barriers to mass production. 

After providing Iris with a research paper about machine learning, it identified and grouped 422 papers by concept.

Iris provides a relevance score of an identified paper.

2. Predicting optimal chemical ingredient combinations: As we’ve discussed with carbon removal devices, machine learning can be used to determine what combination of ingredients yields an optimal outcome. This is relevant for clean meat because the process of growing meat from a cell involves submerging it into a soup of nutrients that allows the cell to grow. Machine learning can help predict the optimal chemical ingredient combinations of that nutrient soup.

Responses to Covid-19

Last on the list today is Covid-19. Covid-19 has killed over 5 million people since it was discovered in 2019. Few priorities are greater than preventing this disease from killing 5 million more people. 

The following are a few ways machine learning is being used to help advance innovation in the response to Covid-19: 

Photo by CDC on Unsplash

1. Understanding the patterns of viral spread: Machine learning can be used to analyze large-scale data on covid-19 patients to identify patterns of spread (5). Machine learning models can be trained to predict what jurisdictions are likely to become a hotspot by reviewing huge amounts of data, including previous spread, demographics of the area, and viral loads found in the sewage system, among other variables. Armed with those predictions, public health officials can target scarce resources accordingly.
2. Developing novel, effective therapeutic approaches: MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Jameel Clinic for Machine Learning in Health used machine learning to discover new drug combinations to treat Covid-19, including Remdesevir (6). The goal of their work was to find a combination of drugs that would quickly address Covid-19 despite the speed and novelty of the disease. In order to do this, they used machine learning models to learn the interactions between drugs and biological targets (molecules and proteins) and the relationships between biological targets and diseases. This allowed them to better understand the interaction between drugs and diseases. These insights led them to the discovery of  new drugs like remdesivir (FDA approved to treat covid-19).
3. Survival predictions for severe Covid-19 patients: Based on clinical data from Wuhan, scientists developed a machine learning model to predict the survival rate of severe Covid-19 patients (7). Training data for these types of models would include which patients survived and which patients did not, personal medical histories, family medical histories, demographic data and more. These predictions could help doctors determine how to identify patients that have a high risk of not surviving and how to allocate resources. 

Sources: 

1. https://www.un.org/en/climatechange/science/key-findings
2. https://arxiv.org/abs/1906.05433
3. (https://www.osti.gov/servlets/purl/915602).
4. https://www.technologyreview.com/2021/04/26/1023636/sustainable-meat-livestock-production-climate-change/#:~:text=Nevertheless%2C%20the%20world%20is%20expected,where%20incomes%20are%20steadily%20climbing.
5. (https://journals.physiology.org/doi/pdf/10.1152/physiolgenomics.00029.2020)
6. (https://news.mit.edu/2021/deep-learning-helps-predict-new-drug-combinations-fight-covid-19-0924)
7. https://www.medrxiv.org/content/10.1101/2020.02.27.20028027v1