Biologics Brief: E5: What’s the Best Antibody Discovery Platform? It Depends on the Biology
Published: September 24, 2026 | Category: Insights, Biologics Discovery
It’s one of the most common questions in antibody drug discovery: what’s the best platform? Hybridoma? Phage display? Single B-cell? Transgenic mice? AI? It’s a reasonable question, but experienced discovery teams rarely start there. They start with a different question: what biological problem are we trying to solve?
In this episode of The Biologics Brief, Mosaic’s Chief Strategy Officer Tracey Mullen sits down with Lippy Lippincott, Chief Strategy Officer at AbTherix, for a candid conversation on how discovery decisions are actually made, what gets overlooked, and where the industry is heading.
Start With the Biology, Not the Platform
The most consistent theme across experienced antibody discovery teams is that platform selection is a consequence of understanding the biology well, not the starting point. Before any technology discussion begins, the first step is defining the target product profile: what does success actually look like for this program?
That means asking questions like: What kind of target are we dealing with? Is this intended to become a therapeutic? Are there species requirements? Is there a known epitope? What is the timeline and budget? What are the first go/no-go decisions the team needs to make?
Only once those questions are answered does the technology conversation become meaningful. And sometimes that conversation leads somewhere unexpected. Teams that arrive convinced they want a particular platform often find, twenty minutes into a biology discussion, that a completely different approach makes more sense. That is not a failure of preparation, it is exactly how good discovery decisions get made.
Budget is also a real constraint that shapes platform choice. Understanding what is being sacrificed at different budget levels is part of the conversation, not an afterthought. The goal is to make sure teams understand what they could be getting and what they will be getting, and make that trade-off consciously.
No Platform Wins. They All Still Exist for a Reason.
One of the most persistent misconceptions in antibody discovery is the tendency to rank platforms as though one is universally superior. It is not a useful exercise. Exceptional antibodies have come out of hybridoma campaigns, phage display, single B-cell workflows, and transgenic mice. Herceptin, one of the most successful therapeutic antibodies ever developed, came from a hybridoma campaign with just nine positive clones.
The better question is not which platform wins. It is what does this platform allow me to accomplish that another one does not?
Every major innovation in antibody discovery has expanded the scientific toolbox rather than replacing what came before. Hybridoma transformed the field first. Phage display unlocked capabilities that were not possible before. Single B-cell technologies dramatically accelerated sequence recovery. Transgenic mice made fully human antibody discovery far more accessible. AI is now layering on top of everything. And crucially, all of those technologies are still here, still generating value, still solving different problems.
The fact that they all coexist is telling. Each one is suited to a different set of conditions, constraints, and biological questions.
“Fully Human” Doesn’t Mean What Everyone Thinks
The terms humanized, fully human, and fully human variable regions get used interchangeably in the industry, but they mean different things, and the distinction matters.
Humanization means taking an antibody that originated in a non-human species, mouse, rat, rabbit, and engineering amino acid changes to make it look more human, with the goal of reducing immunogenicity. The starting molecule was not human.
A fully human variable region is different. When a transgenic mouse encodes human immunoglobulin genes, the variable domains it produces are genuinely human in sequence, not engineered to look human, but derived directly from human germline sequences. No humanization step is required. The final product, reformatted with a human Fc, is a fully human antibody in the same sense as an antibody that might come from a human donor.
This distinction has practical implications. Skipping the humanization step removes a meaningful source of time, cost, and risk from the development path.
Transgenic Platforms Are Not All the Same
Just as platforms should not be treated as interchangeable, transgenic mouse platforms are not all equivalent either. There are specialized models for fixed light chains, long CDR3s, nanobody discovery, hyperimmune formats, and more. Even among platforms designed for standard monoclonal antibody discovery with high human heavy chain diversity, there are meaningful differences in immune performance, repertoire capture, and how well the human antibody loci integrate with the mouse immune compartment.
Regeneron’s VelocImmune design, introduced in the 2000s, was a significant advance, using chimeric human-mouse antibodies where the variable regions are fully human but the constant regions remain murine, which allowed the mouse immune compartment to behave more like a normal non-transgenic animal. Most of the robust transgenic platforms available today use some variation of that principle. But the implementation details vary, and those details affect performance.
Freedom to operate is another dimension that deserves attention earlier than it typically gets. The platform a team chooses today can have implications for intellectual property, licensing, and commercialization years down the road. It is worth understanding the IP landscape before discovery begins, not after a lead candidate has been identified.
Where AI Actually Fits Today
AI is not a standalone discovery platform, at least not yet. The more accurate framing is that AI has the potential to make every discovery platform smarter. It is already showing up across the workflow: target selection, library design, sequence analysis, developability prediction, antibody optimization, and downstream engineering.
Where AI is already creating real value is in optimization, affinity maturation, pH-dependent binding engineering, developability liability removal, multi-parameter design. It enables tighter design cycles and the ability to feed both positive and negative data back into successive rounds, often with fewer iterations than traditional display-based approaches.
De novo antibody discovery fully in silico remains an area of active development and significant hype. The honest assessment is that it is not there yet for most applications. Make-test cycles are still required. The claims of some AI companies are running ahead of demonstrated results. That said, the work coming out of labs focused on computational protein design, including advances in mini-binder design, suggests that meaningful progress is coming. The timeline is uncertain, but the direction is not.
What AI genuinely does is increase the value of good experimental biology. Better upstream data means better models. The two are not competing, they are interdependent.
Diversity and Developability Both Matter, From the Start
A large hit count is not the right metric for a successful discovery campaign. What matters is finding the right antibodies, and that requires both sequence diversity and early developability thinking.
Sequence diversity matters because epitope diversity matters. In many programs, the antibody that performs best in an efficacy model is not the one with the best binding affinity or the cleanest early assay profile. It is the one that hits the right epitope. Without diversity, you may never find it. The magic epitope is often not known in advance.
But diversity alone is not enough. Developability considerations, aggregation propensity, viscosity, self-association, polyreactivity, sequence liabilities, need to be incorporated early, not treated as a late-stage filter. The teams that consistently advance successful therapeutics are the ones thinking about developability from the beginning of the screening funnel, not after they have already committed to a lead.
Nobody gets credit for discovering a molecule that cannot be manufactured, formulated, or developed. Discovery does not end with finding a binder. That is where the next phase begins.
The Question Worth Asking First
Before choosing a discovery strategy, the most important question is simple: what does success look like?
The platform conversation follows from there. Once the target product profile is defined, the biology, the species requirements, the functional goals, the timeline, the downstream decision points, the right strategy often becomes much clearer. And when it does not, that is a signal to keep asking questions before committing to a workflow.
View and listen to more episodes on our podcast: The Biologics Brief.
Speak to a scientist
Connect with our discovery experts to discuss your program and explore how Mosaic can help move your ideas forward.