Single Cell Multiomics and Mechanical Minds
A towering wall of data stands between humankind and a revolution in our understanding of biological systems – but advanced omics, AI, and machine learning can help knock it down
With Noam Solomon, CEO and co-founder of Immunai
What is single cell multiomics – and why is it so exciting for drug discovery?
To understand the function of immune cells and their role in disease and treatment, researchers need to look at genomic, epigenomic, transcriptomic, and proteomic data layers because unique information lies in each. Furthermore, our ability to synthesize and integrate data from each layer is essential to gain a robust understanding of the immune system.
By leveraging approaches such as CITE-seq (developed by Peter Smibert, Immunai’s vice president of functional genomics) and DOGMA-seq (developed by Smibert and Eleni Mimitou, Immunai’s associate director of single cell technologies), researchers can measure the full RNA and DNA in a cell to learn about the degree of genes being expressed, as well as many of the surface proteomes outside the cell. By first isolating these compartments of the cell, then analyzing each of the reactions and interplay between these compartments, we get the full picture of how each and every cell works.
And that’s important for drug development because disease starts at the single-cell level; we need to know where, why, and how cells change to understand and treat disease. Researchers armed with these multiomic capabilities are able to effectively determine how specific therapeutics – and the addition or deletion of specific genes – impact the overall structure and function of a cell. In effect, they can develop therapeutics more quickly, and better understand which drugs will be most beneficial to which patients.
How is machine learning influencing your work?
The overall goal of multiomic single-cell work is to synthesize information from the genome, epigenome, transcriptome, and proteome to derive useful insights on the function and states of cells. But there’s a problem – these sequencing techniques generate data on not thousands, but millions of cells. As data volume swells from a trickle to a flood, the challenge of drawing meaning to the surface escalates. This is where machine learning can help by closing the gap between the noise and the insights by using computational biology to interpret the raw data.
By training proprietary neural network models on the data and leveraging cross-platform translational learnings, it is possible to understand immune responses and create immune profiles based on differentiated elements, such as state-specific expression versus normal expression. These immune profiles can then enable identification of subtle changes in cell types and how they interact with other cells and proteins.
At Immunai, we combine single cell multiomics with machine learning algorithms to profile the immune system. In fact, we have collated the largest database in the world for immunological data – the Annotated Multi-omic Integrated Cell Atlas (AMICA). Using the insights gathered there, our partners have been able to discover biomarkers, validate therapeutic targets, and measure the efficacy of therapies in clinical trials.
With our technology, Baylor College of Medicine, for example, was able to genetically engineer natural killer T cells for CAR-NKT therapy for neuroblastoma.
In another study, Tel Aviv University worked with us to develop a COVID-19 cocktail by identifying and characterizing six antibodies that target and bind to the receptor-binding domain (RBD) of the spike of the virus, neutralizing it.
What are the overriding trends in drug discovery right now?
The pharmaceutical industry loses tens of billions of dollars each year on failed clinical trials and ineffective drugs; companies need to improve the R&D process. Artificial intelligence and machine learning are growing increasingly popular in the scientific community and could help us to understand diseases in new ways, leading to new treatments. Pharma has traditionally run experiments in a dish or in animals, so the results do not always translate well to humans. With a greater volume of biological data, AI can provide a partial alternative that allows us to predict (without actual experiments) the impact of compounds in different biological systems. By adding AI to the mix, researchers can identify drug targets and rule out ineffective options quickly, thus expediting the clinical trial process.
And the biggest breakthroughs in immune system research in recent years?
One recent breakthrough that has already seen success is the development of checkpoint inhibition for cancer patients. The human immune system has multiple braking mechanisms that help protect the host from being attacked by itself, helping control inflammation and autoimmunity; however, it becomes problematic in cancer. By exploiting these checkpoints, tumor cells evade the immune system and survive.
After years of research, we can now block these inhibitory signals to treat multiple types of cancer. Although not all patients respond to checkpoint inhibitors, we are confident that we can improve the response rate by better understanding how the immune system works.
Another field that has blossomed in the past decade is research on the microbiome. It has become clear that the microbiome is one of the most profound environmental factors that influences our metabolism, intestinal health, anti-tumor immunity, and even neurological disorders.
In addition, there has long been consensus that the brain is an “immunologically privileged” site, kept separate from the immune system to protect the brain. It was thought that this was – for the most part – due to a lack of lymphatic drainage. Yet, a few years ago, scientists found an unexpected lymphatic network in the central nervous system. Given that the activation of the adaptive immune system begins in our lymphoid organs, this discovery will serve as one of the major stepping stones in a leap into a new era in the study of neurological diseases. Being able to understand how the brain interacts with the immune system at a granular level will undoubtedly speed up therapeutic discovery for neurological diseases.
What else must scientists do to push the field further?
When you look at two disciplines we employ in our work – neural networks and cancer immunology – you may realize that these were both initially dismissed as ineffective. Today, not only are these some of the most exciting areas of investment within their respective industries, but the synergies between the two are advancing previously unimaginable cures for disease. Within scientific research, we’ve become accustomed to seeing immediate results, so it is important that we don’t dismiss theories when they don’t immediately work as expected. We need to continue experimenting across different fields, because the answers to the world’s most pressing questions may already exist – though perhaps in another corner of academia.
In addition, we need to recognize that better inputs produce better outputs. As we advance our methodologies for cell sequencing, we can feed better data to our AI and machine learning programs – thus advancing our understanding of the underlying mechanisms of biology.