Driving Drug Discovery Success With Synthetic Biology and NGS
Industry Insight Jul 15, 2020 | By Laura Elizabeth Lansdowne, Senior Science Writer, Technology Networks.
Understanding protein–protein binding is critical for the development of antibody-based therapeutics. Typically, when treating diseases such as cancer, the aim is to bind to a single protein target. However, when it comes to infectious diseases, the therapeutic target can consist of many related but distinct proteins.
Technology Networks recently spoke with David Younger, PhD, Co-founder and CEO of A-Alpha Bio, to learn more about how the company has harnessed synthetic biology and next-generation sequencing techniques to measure millions of interactions between proteins in a single test tube. This enables the identification of potential drugs for the treatment of infectious diseases, including COVID-19.
Laura Lansdowne (LL): How can synthetic biology and next-generation sequencing (NGS) be harnessed to drive drug discovery success?
David Younger (DY): Synthetic biology and NGS are powerful tools that can be harnessed to generate enormous amounts of data. Engineered cells can act like microscopic test tubes – each one representing a unique condition to test a specific hypothesis. Many cells can be pooled together to perform millions of experiments in one normal sized test tube. With an enormous number of experiments mixed together, NGS allows us to measure each outcome separately and to associate it with a particular experiment.
At A-Alpha Bio, we focus on experiments that measure interactions between proteins. Understanding how proteins interact with one another is critical for the development of protein-based drugs, including antibodies, that function by binding to a disease-causing target protein. Antibody discovery and optimization involves testing for binding between many candidate antibodies and many target proteins to evaluate the effectiveness and safety of each antibody. Our AlphaSeq platform utilizes synthetic biology and NGS to measure millions of interactions between proteins in a single test tube.
The development of the AlphaSeq platform began with a very simple observation – proteins coating the surface of two different types of cells act like Velcro and cause the cells to stick. We figured out how to shave off the natural protein “Velcro” and replace it with therapeutically relevant antibodies and antigens. We also figured out how to use next-generation DNA sequencing to measure cell–cell sticking. The result is the AlphaSeq assay – we can combine thousands of engineered cells with unique protein coatings and determine which bind to one another and with what strength.
LL: How does A-Alpha Bio’s approach differ to other existing technologies used to identify promising drugs for the treatment of infectious disease?
DY: Information about protein–protein binding is critical for the development of any antibody drug, but there is an additional complication with infectious diseases. For most antibodies, the goal is binding to a single target, such as a receptor present on cancer cells. For infectious diseases, the “target” consists of many related but distinct proteins. The best-known example is probably influenza. Each season, this virus randomly mutates to make the previous year’s vaccine ineffective and emerges as multiple circulating strains. Antibody drugs against flu have failed because they are too specific to a particular strain. Instead, infectious disease drugs require cross-reactivity – or the ability to bind to many different strain variants.
AlphaSeq is unique in that it allows us to measure the binding of many antibodies against thousands of different targets at once. Currently, drug developers identify an antibody that binds to one viral or bacterial strain and then test the antibody for binding to other strains one at a time. This process is often prohibitively slow and costly to discover and optimize cross-reactive antibodies. Using AlphaSeq, we can map a complete interaction network between many antibodies and many target variants to find those that exhibit cross-reactivity. We can also use the resulting data to train machine learning models to further accelerate the development of potent and effective drugs.
LL: Can you elaborate on how the AlphaSeq platform will be used to aid development of potent and cross-reactive therapeutics against intestinal and respiratory pathogens?
DY: We have previously shown that we can use AlphaSeq to measure binding between a panel of antibodies and a panel of protein targets from intestinal and respiratory pathogens. With continued support from the Bill & Melinda Gates Foundation and a collaboration with Lumen Bioscience, we will apply the AlphaSeq platform to develop and optimize potent and cross-reactive antibody therapeutics. We will build large panels of antibody candidates and panels representing variants of the infectious disease targets. We will then use AlphaSeq to map interactions between the antibodies and target proteins and determine the strength of each protein–protein interaction. This data will be used to train a machine learning model that will predict antibodies with improved potency and cross-reactivity. Multiple iterations of machine learning and AlphaSeq will be performed to identify promising lead antibodies for continued development.
LL: What intestinal and respiratory pathogens are of particular interest?
DY: We expect that AlphaSeq, combined with machine learning, will be a valuable tool for developing antibodies targeting a wide range of infectious diseases that affect both developed and developing countries. Initially, we will apply AlphaSeq to optimize antibodies against two very different pathogens: Campylobacter, a bacterial pathogen that causes diarrheal disease, and SARS-CoV-2, the virus responsible for COVID-19.
LL: How can data generated using the AlphaSeq platform be used to develop machine learning (ML) models? Can you touch on the benefits of using ML in drug discovery?
DY: ML requires training datasets to build models that have predictive power, and the larger and higher quality the training data the better the model will be. For the optimization of antibody potency and cross-reactivity, the most valuable data is quantitative and involves binding measurements between many antibodies and many target protein variants. AlphaSeq is perfect for generating this type of data. We can include thousands of antibody variants and hundreds of pathogen variants in a single iteration of the AlphaSeq assay. ML can then be used to determine the rules that drive potency and cross-reactivity to each of the target variants and predict optimized antibodies.
In general, ML has enormous implications for drug discovery. Biological systems are incredibly complicated and useful trends and patterns are often hidden in the complexity. In the case of AlphaSeq, we can take a thousand antibodies and a thousand antigens and map all of the million protein interactions between each antibody and each antigen. This is an enormous amount of data – too much for a scientist to manually go through each interaction and try to understand the underlying rules that allow a particular antibody to bind to a particular antigen. With ML, however, we can do exactly that. If synthetic biology and NGS allow us to produce and measure an unprecedented amount of data, ML allows us to efficiently interpret and apply that data – for example, to develop more effective drugs for infectious diseases.
David Younger, PhD., was speaking with Laura Elizabeth Lansdowne, Senior Science Writer for Technology Networks.