Antibodies have been central to the biologics sector, with over 160 monoclonal antibodies approved by the FDA and accounting for a significant portion of top-selling drugs. However, protein engineering is expanding beyond traditional antibody formats as new modalities such as bispecific antibodies and antibody-drug conjugates (ADCs) gain ground. As of 2024, there are more than 200 ADCs in clinical development and 19 bispecific antibodies approved worldwide, generating over $12 billion in sales.
With regulatory agencies like the FDA increasingly supporting novel therapies in 2025, the focus within drug discovery has shifted from identifying targets to addressing the challenges of producing and scaling complex molecules. Companies such as Viva Biotech, a contract research organization specializing in biologics and structural biology, report growing demand for next-generation formats including bispecifics, ADCs, and other innovative antibody programs.
Paul Wan, vice president of early discovery and business development at Viva Biotech, described a major challenge: “The biggest problem with these biologics is really the heterogeneity in actually generating these molecules. You’re getting mispaired chains in bispecifics or trispecifics and also different conjugation variants when you’re looking at ADCs.”
Industry experts note that inconsistent drug-to-antibody ratios (DARs) have contributed to clinical delays for ADC candidates. In response, platform-level engineering techniques—such as controlled Fab-arm exchange, knobs-into-holes methods, crossMabs technology, and site-specific mutations—are being used to standardize molecule assembly. Advances in conjugation chemistry now allow for better control over DARs by incorporating non-natural amino acids.
Artificial intelligence (AI) and machine learning are increasingly integrated into early-stage screening processes to predict developability issues such as aggregation or glycosylation before manufacturing scale-up begins.
Automation has accelerated protein production workflows. Laboratories are using parallel purification systems that enable simultaneous expression and screening of hundreds or even thousands of constructs—a level of throughput not possible a decade ago. Industry analysis indicates that adoption of novel formats increased by 30% during the first half of 2025; automation has allowed labs to achieve up to five times faster construct turnover rates.
Jerry Zhang, director of biology at Viva Biotech explained: “High-throughput protein production can allow us to try different variations, and potentially it can help us to find the best construct that we are looking for that we can use for protein production or some structural biology or even some assays.”
AI’s role has expanded from structure prediction to encompassing all aspects of protein engineering—including sequence analysis and manufacturability assessment—with deep learning models enabling more accurate predictions about stability and solubility. Yue Qian, executive director of computational chemistry and artificial intelligence platform at Viva Biotech said: “Conventionally, there are several different ways to extract protein features and to present these proteins. These are mostly sequence-based. But nowadays, since we have protein large language models, they provide much richer approaches to better describe these proteins. And on top of that is the recent advance of the AlphaFold-like structure prediction tools.”
Qian added: “We’ve already seen publications in the past couple of years talking about how they can improve GPCR stability and solubility. And we’ve also seen pretty accurate models to predict the protein yields based on the sequence.” AI-driven structure prediction accuracy has improved substantially over two years according to industry reports.
Viva Biotech employs an AI-driven drug discovery (AIDD) platform consisting of three components: V-Scepter for parameterization; V-Orb for physics-based modeling; and V-Mantle for generative design—all intended to bridge concept development with manufacturability.
Despite progress made through AI integration with wet-lab validation data—including both successful outcomes and failures—challenges remain regarding membrane protein expression consistency across different systems.
Beyond antibodies, researchers are developing new classes such as PROTACs (proteolysis-targeting chimeras), molecular glues that induce targeted degradation via novel interactions between proteins inside cells; more than 120 PROTACs plus around 60 molecular glues are currently advancing through preclinical or clinical stages globally by 2025.
Wan noted advances in linker library generation (“linkology”) which supports rapid reaction screening without purification steps using direct-to-biology platforms; biosensors employing FRET technology assist with ternary complex validation while CRISPR-based cellular assays track real-time degradation events inside cells.
Molecular glues pose unique challenges due their single-molecule mechanism but can be screened using DNA-encoded libraries coupled with affinity selection mass spectrometry despite lower hit rates compared with traditional binders.
Another emerging area involves de novo mini-protein design powered by AI-guided modeling combined with advanced structural biology techniques like cryo-electron microscopy (cryo-EM).
Qian highlighted ongoing efforts at Viva Biotech integrating computational design with experimental optimization: “Protein engineering needs not just better models but data that truly reflects biology.”
A major barrier cited across industry is lack of standardized datasets—making benchmarking difficult—which public-private consortia aim to address via principles like FAIR (findable, accessible, interoperable reusable).
Looking forward industry leaders expect further integration between computational predictions experimental feedback anchored by high-resolution structure-based drug design methods will drive reliable advancement complex biologics from laboratory into clinic.
This article was written in partnership with Viva Biotech.
