Reinventing the Discovery and Cell Line Development of Biotherapeutics

Episode Summary

Executive Summary: This BioReport episode features AbSci CEO Sean McLean explaining how the company uses AI and synthetic biology to collapse biologics discovery and cell line development into one workflow. By screening billion-member libraries in E. coli, integrating patient-derived target discovery, and feeding results into deep learning, AbSci aims to shorten timelines, improve manufacturability, expand therapeutic possibilities, and lower costs.

Main Topics: Collapsing discovery and cell line development (Priority: 5/5): AbSci’s core thesis is that drug discovery and manufacturing should not be separate steps. The platform simultaneously evaluates affinity, titer, and quality to avoid reformatting and cell-line transfer failures. AI-driven biologics design (Priority: 5/5): The company uses deep learning models trained on experimental data to predict optimal protein sequences, library design, and associated manufacturable cell lines for specific targets. Patient-sample target discovery via Totient (Priority: 4/5): After acquiring Totient, AbSci can use patient-derived samples and bulk RNA-seq to identify antibodies and novel targets, then pan those antibodies against the proteome to advance new programs. Scale and selection of protein libraries (Priority: 4/5): AbSci screens billion-member libraries, then uses flow cytometry and multi-stage down-selection to identify top candidates that can scale to GMP manufacturing. E. coli as a manufacturing and screening advantage (Priority: 4/5): The platform uses E. coli to accelerate cycle times, simplify engineering, and reduce cost of goods relative to CHO/mammalian systems. Bionic proteins and site-specific bioconjugation (Priority: 3/5): AbSci’s non-standard amino acid platform enables homogeneous, site-specific conjugation for applications such as pegylation and ADCs, addressing incorporation and titers issues. Business model and use of IPO proceeds (Priority: 3/5): AbSci remains a pure-play technology company, partnering with pharma/biotech and earning milestones/royalties while using IPO capital to expand R&D, talent, and AI capabilities.

Key Arguments: Traditional biologics discovery fails at two points: molecular reformatting and switching from transient to stable cell lines; AbSci claims to eliminate both by doing discovery and cell line development simultaneously. Training deep learning models on affinity, titer, quality, and target sequence data can enable prediction of not just a best candidate sequence, but also the best manufacturing cell line. Screening billion-member libraries in the exact patient-facing format is a major leap versus the thousands or tens of thousands typically screened today. E. coli offers faster fermentation, easier engineering, and materially lower cost of goods than CHO or mammalian systems, making it better suited to AbSci’s workflow. Acquiring Denovium brought deep-learning capability in-house and has already produced measurable gains, including a 2x titer improvement from a model-predicted strain. Patient-derived samples may reveal different immune responses than conventional B-cell single-cell sequencing, enabling discovery of novel antibodies and targets from tissue-specific biology. The platform’s long-term goal is to become the 'Google' or 'Google index search' of protein-based drug discovery, eventually moving fully in silico. AbSci’s partnership model lets it avoid the binary clinical risk of developing drugs itself while still sharing in upside through milestones and royalties.

Data Points: Library size: 1 billion-member library - AbSci screens massive genetically distinct variants in its discovery and development workflow Candidate down-selection: Top 1,000 hits - Initial billion-member screen is narrowed for deeper characterization and manufacturability checks E. coli fermentation time: 2 days - Used to illustrate speed advantage over mammalian/CHO systems Mammalian/CHO fermentation time: 14 to 21 days - Comparison point for the platform’s faster bacterial workflow Cost of goods reduction: 50% to 75% - Estimated benefit from using E. coli in the cell line development process Titer improvement: 2x - Result from a strain predicted and tested using the integrated deep learning model Non-standard amino acid incorporation: High incorporation rates - Claimed improvement enabling bionic proteins and site-specific bioconjugation Patient sample source: Lung fluid samples from COVID patients - Totient example used to identify neutralizing antibodies during the pandemic Neutralizing antibodies discovered: 15 - Result of Totient’s pandemic-era lung-fluid sample work IPO proceeds: About $200 million - Raised in July to expand capacity, talent, and R&D

Pivotal Quotes: "We are able to discover the drug candidate at the exact same time we're developing the cell line." — Sean McLean: Explaining AbSci’s central approach of collapsing discovery and manufacturing into one process "Our vision is to become the Google index search of protein-based drug discovery and biomanufacturing." — Sean McLean: Describing the long-term ambition for AI-driven protein design and manufacturing "We're able to predict the absolute best drug candidate for a particular target." — Sean McLean: Discussing the end-state for the company’s deep learning platform

Implications: If AbSci’s platform generalizes, biologics development could become faster, cheaper, and more predictive, with earlier manufacturability built in. That could expand the set of druggable proteins and improve success rates for partners.