When Antibody Purification Hits an 'Exception' — The Limits of Platform Processes and How to Handle Them

In the biopharmaceutical field, the purification process of monoclonal antibodies is recognized as one of the most mature platforms. In 2023, the global market for antibody drugs was estimated to have reached around $200 billion, making up half of the entire biopharmaceutical market. From cancer immunotherapy to autoimmune diseases, from neutralizing antibodies for infectious diseases to antibody-drug conjugates (ADCs), these Y-shaped glycoproteins have become a core weapon in modern medicine.

However, there's one number in the cost structure of antibody drugs that rarely gets public attention: according to several industry reviews, downstream purification costs typically account for 50% to 80% of total production costs. This means that more than half of the manufacturing cost for each antibody drug is spent on "separating the antibody from the cell culture." When a product’s annual sales reach tens of billions of dollars, every percentage point in purification efficiency can translate to tens of millions of dollars in cost differences.

This is antibody purification—a part of the biopharmaceutical industry that doesn’t directly create therapeutic effects, yet profoundly impacts whether those effects can reach patients at an accessible price.

Protein A: One Protein Defines an Industry

If there’s one protein that has shaped the modern antibody industry, Protein A is the most frequently mentioned. This cell wall protein from Staphylococcus aureus has an important trait: it can specifically bind to the Fc region of most IgG subtypes with nanomolar affinity, and it can be released reversibly under low pH conditions.

This feature was turned into a chromatography medium in the 1970s, profoundly changing the technical path of antibody purification. Before this, antibody purification relied on multi-step processes like ammonium sulfate precipitation, ion exchange, and gel filtration, each step causing yield loss. The emergence of Protein A allowed for a single step of affinity capture to achieve over 95% purity, anchoring the entire downstream process at a highly selective starting point.

Protein A

Nowadays, the vast majority of commercial therapeutic antibodies—including IgG1, IgG2, and IgG4—use Protein A affinity chromatography as the first step in the downstream purification platform. The success of this platform has made antibody purification somewhat 'standardized': cell culture feed → Protein A capture step → low pH viral inactivation → ion exchange polishing (AEX/CEX) → virus filtration → ultrafiltration/diafiltration. This workflow has been widely replicated in the industry.

However, the flip side of platformization is a systematic neglect of ‘exceptions.’

Achievements and Costs of Platformization

Platformization brings efficiency, predictability, and regulatory convenience to antibody purification. Regulatory agencies recognize validated platform processes, technology transfers have clear references, and process development teams can usually lock down a preclinical process for a new IgG1 molecule within 6 to 9 months.

But the premise of platformization assumes that all antibody molecules 'behave similarly.' In reality, this isn’t the case.

Different IgG subtypes have significant differences in their affinity for Protein A. IgG1 binds strongly and stably; IgG3, however, has weakened hydrophobic interactions and hydrogen bonding with Protein A due to key amino acid replacements in the Fc region’s CH3 domain (like His435→Arg435), resulting in significantly reduced binding. Some IgG3 allelic variants may not bind Protein A at all. Even within the same subtype, differences in the variable region sequences—especially certain VH3 family germline genes—can introduce additional Protein A binding sites, altering elution behavior. Individual mutations in the framework region may seem to have no effect on antigen binding, but can significantly change antibody thermal stability and aggregation tendency.

When these molecular 'personalities' accumulate in industrial-scale operations, platform processes start to show cracks. A purification template that works well for IgG1 may see a sharp drop in elution recovery, excessive aggregates, or unusually high host cell protein (HCP) levels when applied to an IgG2 or bispecific antibody. This isn’t necessarily a failure of the platform process itself, but more likely reflects an inherent tension between platform thinking and molecular individuality.

Elution: The Key Engineering Lever

If Protein A capture is the starting point of antibody purification, elution is the critical step that determines success or failure.

The standard approach for Protein A elution is low pH—usually between pH 3.0 and 3.5. In an acidic environment, the hydrophobic interaction between the antibody Fc region and Protein A weakens, and the antibody releases from the medium. But this pH range is precisely where antibodies are more prone to conformational changes and aggregation. Low pH-induced aggregation is a major reason for yield loss and product quality issues in antibody purification.

It’s even more complicated because the 'optimal elution pH' is a parameter that varies from molecule to molecule. For IgG1, the typical elution window is pH 3.2-3.5, but some molecules start aggregating at pH 3.5, while others don’t fully dissociate even at pH 3.0. There are reports that lowering the elution pH by 0.2 could increase the aggregate level from 2% to 15%—this difference might just show up as a small peak on a chromatogram, but it’s enough to trigger a deviation investigation in product quality reviews.

Besides pH, the type of buffer salts, ionic strength, and the addition of stabilizers like arginine all make up the multi-dimensional parameter space for elution conditions. In process development, a common approach is to run gradient experiments from pH 2.8 to 4.0 in 0.2 pH unit steps, and add arginine or urea as stabilizers if needed. This takes time, and each round of testing consumes cell culture supernatant and chromatography media.

Here’s a logic that’s easy to overlook: the goal of optimizing Protein A elution isn’t necessarily to maximize yield at this step, but to balance its overall impact on the final product quality. The globally optimal elution pH usually sits near the inflection point of the 'yield-aggregate' curve, not at the point with the highest yield. For example, if an antibody already has more than 95% elution at pH 3.5, dropping the pH further to 3.2 might only improve yield by less than 2%, but could push aggregate levels from 2% to 6%—these extra aggregates will need to be partially removed in downstream purification, and each purification step comes with its own yield cost. This logic aligns with the process development principle that 'the sum of local optima doesn’t necessarily equal the global optimum.'

Aggregates: a quality risk that’s easily underestimated.

Antibody Aggregation and SEC Removal Challenge

One often overlooked challenge in antibody purification isn’t low yield, but controlling aggregate content. Aggregates are non-covalent or covalent oligomers of antibody molecules. Small amounts of dimers and trimers might overlap with the monomer peak on purification profiles, making them hard to accurately detect with routine analyses. But once they’re present in the final product, they can cause immunogenic risks, reduce efficacy, or affect drug pharmacokinetics. Regulators are becoming stricter about aggregate content, typically requiring high-molecular-weight aggregates in the final product to be below 1%-2%.

The mechanisms behind aggregate formation are quite complex: they might result from low pH exposure during purification, molecular crowding at high concentrations, or shear stress. The tougher part is, once aggregates form, they’re hard to remove. Preparative size-exclusion chromatography (SEC) is one of the main ways to remove them, but it has limited sample capacity, takes a long time, and usually has yields under 80%, making it one of the more costly steps in the purification process.

So a more effective strategy to control aggregates is often not 'remove them after they form' but reduce their formation from the source. This requires understanding each antibody’s unique aggregation tendencies—is it hydrophobic interactions between VH domains? Flexibility in the hinge region? Or conformational changes in a certain CDR loop under low pH? In traditional platform development, this information usually only gets attention at pilot-scale production.

When platforms meet exceptions

The limits of platform processes are especially evident with new types of antibody molecules. Bispecific antibodies, engineered to bind two different antigens, are usually structurally more complex—with asymmetric chain pairing, extra domains, and non-natural linkers—their low-pH stability and aggregation behavior often differ noticeably from natural IgGs. ADCs have small-molecule toxins attached to the antibody backbone, and the toxin load changes the surface hydrophobicity and charge distribution, directly affecting non-specific adsorption and recovery during purification. Antibody fragments (Fab, scFv, nanobodies) lack the Fc region and don’t bind Protein A, meaning the assumptions of the entire platform workflow no longer hold.

The proportion of these 'non-standard' antibodies is rising rapidly. It’s estimated that more than 30% of antibodies in clinical stages are non-traditional IgG1 formats. The coverage of platform processes is being outpaced by the growth of these new molecules.

Making exceptions predictable

When a field starts running into 'exceptions' frequently, it usually means the existing experience framework needs to expand. MatwingsVenus™ (Xiaowu™) intelligent agent’s capabilities provide computational support for this expansion—it’s not about replacing platform processes, but giving the platform the ability to handle a broader range of molecular diversity.

MatwingsVenus™

Pre-classification of antibody subtypes and germline genes. As soon as the sequence is obtained, the MatwingsVenus™ (Xiaowu™) AI can identify the antibody's subtype, germline gene family, and key residues in the framework regions. Differences in the IgG3 Fc sequence, extra Protein A binding sites in the Fab region of the VH3 family, and specific framework residues reported in the literature that are related to low pH stability—this information helps developers make an initial judgment about the molecule's compatibility with standard platforms before experiments even begin and flag parameters that might need adjustment early on.

Rational prediction of elution pH. The protein function prediction module (VenusX/VenusG) can use the sequence to assess the antibody's tendency to aggregate and changes in charge under low pH conditions. Combined with structural predictions (AlphaFold2/ESMFold/Protenix) for visual analysis of hydrophobic patches on the Fc and Fab surfaces, it provides a qualitative risk assessment signal, flags high-risk molecules, and helps design experimental pH gradients. If the prediction suggests that the antibody has a notable risk of unfolding or aggregating below pH 3.2, the elution strategy can prioritize a milder pH range with stabilizers—not something discovered after experiments, but proactively avoided during the design stage.

Molecular-level early warning of aggregation risk. Aggregation tendency is one of the harder-to-predict variables in antibody purification, but it's not completely unpredictable. 3D structural analysis can help identify potential aggregation hotspots—highly exposed hydrophobic residues in the CDRs, flexible hinge regions, and charge imbalances at the VH-VL interface. The correlation between these features and aggregate formation has been reported in multiple studies. When a candidate molecule is predicted to have a high risk of aggregation, the development team can proactively plan countermeasures—optimizing elution pH, adding stabilizers, or scheduling extra SEC purification steps—instead of waiting to adjust passively based on aggregate content measurements at the pilot scale.

Designing 'better-to-purify' antibodies from the start. If purification-relevant molecular properties are considered early during candidate screening, the pressure during later purification development can be reduced. MatwingsVenus™ (Xiaowu™) AI's protein design capabilities (VenusREM/VenusPrime) can recommend candidate framework mutations to potentially improve low pH stability or reduce exposure of surface hydrophobic patches, all without affecting antigen binding. Sometimes a smoother purification path isn't just found during purification steps, but can be strategically laid out during the design stage.

Conclusion

The history of antibody purification is a spiral, moving from diversity to standardization, and then back from standardization to confronting diversity again.

The discovery of Protein A allowed antibody purification to converge from a cumbersome multi-step process into an efficient platform. Over the past thirty-plus years, this platform has supported the rapid growth of the global antibody industry. But the emergence of new molecular formats is gradually making this platform go from being “almost suitable for all molecules” to “suitable for most molecules, but with some notable exceptions.”

MatwingsVenus™ (XiaoWu™) intelligent system’s exploration here isn’t about creating another platform to replace Protein A. It’s about providing developers with more computational reference points when dealing with molecules that deviate from the standard. When key molecular properties can be preliminarily assessed before experiments, the choice of purification routes can shift more from “fit existing templates” to “tailor to specific molecules.”

When exceptions become predictable, the platform is no longer a one-size-fits-all rule but an intelligent system that can dynamically adjust to a molecule’s unique traits. This might be a potential direction for the future evolution of antibody purification.