Unlocking Measurement-driven Drug Design

September 2020

Most drug developers would agree that the inherent complexity of biology makes it incredibly difficult to predict how a molecule will ultimately interact with an intricate system like the human body. The path from molecule to drug requires a series of experimental assays from in vitro to in vivo. As the assays grow in complexity and disease relevance, the number of molecules that can be tested falls from thousands to only a handful. Shepherding drug candidates through this process is a long, expensive, and emotionally trying process, not only for the talented people doing the work, but especially for concerned patients and their families.

Manifold Bio was born from a deep conviction that we could upgrade the toolkit available for creating life-saving drugs. As PhD students in George Church’s lab, Pierce and Gleb innovated with cutting-edge DNA sequencing and synthesis technologies to engineer genomes, viral capsids, and other proteins. We merged Pierce’s cloning wizardry and Gleb’s computational expertise into a seamless process. We built novel assays together and pair programmed to design the DNA sequences going in and analyze the torrent of data coming out. What had attracted us both to George’s lab was his vision to increase the scale at which we could read and write biology. Under his mentorship, we came to see a path to unlocking an exponential increase in measurement at all stages of protein therapeutics development. We made it our mission to transform the pursuit of these medicines from one of discovery to one where molecules with properties previously thought to be impossible are designed.

Over the past year, we set out to get to know folks in the drug discovery community in order to understand the key bottlenecks. Focusing on protein therapeutics, we came to appreciate the challenges faced by diverse teams working together to optimize the properties of a molecule with the shared goal of achieving strong biological readouts. Several patterns emerged:

  1. there is a need for new and better ways of generating data; but more importantly
  2. there is a need for more data from relevant disease models to better predict outcomes in patients; and
  3. there was simultaneous curiosity and understandable skepticism about the promise of machine learning in drug discovery, considering the quantity and quality of data currently available.

At this time, Pierce was completing his PhD thesis using advances in DNA synthesis combined with DNA sequencing to answer key questions in the gene therapy field in a way no one else had before -- testing all 200,000 single mutations to the capsid protein of AAV, in a single mouse. This elegant experiment enabled his team to simultaneously engineer multiple objectives at once, including targeting specific tissues, improving viral packaging, and reducing immunogenicity. This was possible because the identity of a virus mutant could be read by sequencing its DNA, which served as a natural barcode, and all virus mutants from this single library experiment could be quantified in multiplex.

We saw the immense impact that the power of multiplexed testing could bring to the broader class of all protein therapeutics. But there was a key challenge: while AAV variants can be read out directly by sequencing the DNA contained within their genomes, protein therapeutics are not naturally attached to unique DNA signatures that enable economical read-out. The process of chemically conjugating DNA is slow, expensive, damaging to the protein, and limits the kinds of uses Pierce invented with AAV, especially in vivo testing. And, crucially, the equivalent technology for multiplexed sequencing of proteins directly is yet far away.

At Manifold Bio, we are working on a novel approach to overcome this challenge--a protein barcoding platform that will allow efficient tagging and quantification of vast libraries of therapeutic protein designs. We envision a modular, plug-and-play technology of barcodes that can be applied to thousands of protein designs of interest, unlocking the ability to track them, in mixture, across assays, both in vitro and in vivo. We’re enabled to achieve this vision by a $5.4 million investment and complementary expertise in building hard platforms from our seed investors Playground Global and Fifty Years, following early support from GETTYLAB and Allston Venture Fund.

We believe this technology has the potential to accelerate the entire industry. With more throughput, teams can widen their screening funnel to include discovery in vivo, rather than reserving in vivo models for validation of “top” drug candidates. Targets and the molecules to effectively drug them can be discovered through massive in vivo screens, yielding drugs with the right tissue penetration and the right PK earlier in the process. Further, more and better data has another promising consequence: unlocking the true potential of machine learning for drug design. The co-founders at Manifold Bio know from experience that machine learning predictions are only as good as the data that drives them. Manifold’s protein barcoding platform aims to unlock  measurement-driven drug design --learning the principles that determine good drugs from abundant measurements in complex environments representative of where they’ll be deployed.

Our respect for the complexity of biology and our commitment to enabling the drug discovery process shows in the highly technical and application-focused team we’ve built at Manifold Bio so far. Early on, we challenged our co-founder Shane Lofgren to imagine how we could use the platform to find novel targets in cancer, an area in which he has published over a dozen papers as a bioinformatician and helped spin out and co-found companies from Stanford and MD Anderson. We also brought on Kate Nudel, a multi talented cell biologist who led multiple programs at Kintai Therapeutics applying their unique multi-omics platform to a number of disease models. Hoong Chuin Lim is a driven molecular biologist who tamed an entire model organism from scratch. Pierce and Gleb built the foundational infrastructure for Manifold’s novel library-guided protein engineering approach, with next-generation DNA sequencing synthesis and advanced bioinformatics infused throughout. Now, the opportunity for this early team is to leverage this approach to create a platform that will deliver on its promise to bring better drugs more quickly to patients.

Manifold’s protein barcoding platform and the measurement-driven drug design that it will enable are all in service of pursuing a vision of the future of drug discovery: one where our internal teams as well as teams of our partners in the industry are massively more able to achieve the goal we all hold in the field--bringing life saving drugs to patients. If this is a future you want to join us in building, either as our next incredible hire or as an early industry partner, please reach out at We would love to chat.

The Manifold Bio Team