AE&T 2025 felt like the moment “AI-native” antibody R&D became operational, not aspirational. The industry is moving from isolated models to full-stack systems that connect in vitro, in vivo, and in silico workflows into a lab-in-a-loop. The bottleneck is no longer algorithms, it’s data + workflow fragmentation and the platform discipline to fix it.
Key Takeaways:
2025’s theme: embed AI into every layer of antibody engineering, not as an afterthought.
Full-stack discovery connects in vitro, in vivo, in silico into a “lab-in-a-loop.”
The biggest blocker is fragmented data & workflows across ELN/LIMS/SDMS/point tools.
Developability needs to be platform-level and early, especially for complex formats.
The next wave expands beyond oncology into broader indications.
For years, antibody engineering conferences showcased hero molecules and flashy formats. AE&T 2025 felt different—less about individual breakthroughs, more about the digital and AI infrastructure needed to reliably turn complex protein designs into real therapeutics.
Antibody Engineering & Therapeutics 2025 felt like the first AE&T where “AI-native” antibody R&D stopped being aspirational and instead became the organizing principle for how R&D teams think about protein formats, workflows, and platforms.
Underneath the buzzwords, the real story was the uneasy, productive tension between ambition—multispecifics, ADC, novel scaffolds—and the stubborn realities of AI, data quality, developability, tech debt, and organizational change.
What dominated AE&T 2025: Full-stack, AI-native discovery
Across plenaries, talk tracks, and vendor booths, one theme dominated: integrating AI and computational design into every layer of antibody engineering rather than treating it as an afterthought. Talks didn’t just present individual models; they framed full-stack discovery systems that connect in vitro, in vivo, and in silico discovery.
The clear signal is the need to leverage the confluence of biophysical data, structural predictions, and developability analytics into an efficient “Lab-in-a-Loop.”
This shift showed up in multiple topic areas like AI-assisted design of bispecific and multispecific formats, ML-guided developability triage, and data-aware optimization of ADCs and cell engagers.
There was also clear emphasis on moving beyond oncology, extending these tools to fibrosis, cardiovascular, and neurodegenerative disease, which isa shift for AI-empowered engineering toward a general-purpose engine for indication expansion.
What changed vs prior years
Compared with earlier AE&T meetings where bispecifics, ADCs, and immune agonists were the focus, 2025 seemed less like a “zoo” of formats and more like a discussion around the infrastructure it takes to make those formats reliably developable at scale.
Previous years spotlighted “hero” molecules or novel architectures; this year returned to questions of platform, data standards, and how to operationalize computational design inside large and mid-size organizations.
There was also a noticeable evolution of the AI narrative with less “we trained a model and it’s amazing,” and more “here is how we embedded models into our screening funnels, here is what broke, and what actually moved a program metric.” The conference agenda reflected this evolution with dedicated sessions on AI-driven antibody discovery, expanded workshops on developability and candidate selection, and clear framing of computational design as a core theme rather than an adjunct track.
Scale, vendors, and the hallway read
AE&T 2025 hosted solution providers across the discovery, engineering, analytics, and informatics landscape. The exhibition floor reflected a concentration of companies positioning themselves around high-throughput discovery, AI-ready data capture, and specialized services for bispecifics, ADCs, and novel scaffolds.
Anecdotally, the meeting reflected a field in full-on commercialization mode: CROs and CDMOs leaning into integrated “design-to-IND” offerings, software companies emphasizing multimodal data platforms, and instrument providers highlighting how their hardware plugs directly into workflows.
Recurring challenge: data and workflow fragmentation
The most persistent challenge that surfaced in talks, posters, and hallway chats focused on the fragmentation of antibody R&D workflows and the data they produce. Teams are collecting massive volumes of sequence, structure, binding, potency, and biophysical data across in vitro display, in vivo models, and developability screens, but those assets are still siloed across ELNs, point solutions, LIMS, SDMS, and bespoke scripts.
Speakers repeatedly mentioned that AI and computational design only deliver value when supported by curated, context-rich, multimodal data streams. Without that foundation, models either overfit to narrow data sets, fail to generalize to new formats like trispecifics or degraders, or become isolated “experiments” that never impact portfolio-level decisions. The frustration was palpable, with a shared recognition that while the technology is available there’s organizational readiness and data architecture lag to wrestle.
A move from artisanal protein engineering to industrialized intelligence
The frontier of antibody engineering—multispecifics, ADCs with novel payloads, immune agonists, GPCR-targeting scaffolds—is exploding with format design dimensions, risk modes, and failure pathways. As formats become more complex, intuition-driven or single-modality optimization breaks down; the cost of missing early immunogenicity flags or developability liabilities only increases nonlinearly with format sophistication.
Plus, market and clinical expectations keep rising. Payers and regulators are pushing for differentiated mechanisms, better safety profiles, and clearer biomarker strategies, all of which depend on richer, integrated data.
Antibody discovery and development teams need to not just leverage better assays or models; they need end-to-end systems that treat data as a strategic asset to enable continuous learning from past programs. Then feeding those learnings back into design and selection upfront.
What a multimodal scientific intelligence platform needs to unify
I had the opportunity to present on “Empowering Next-Gen Protein Design: A Multimodal Intelligence Platform for Advanced Immunotherapy” a topic squarely within the broader need toward integrated, AI-ready antibody R&D.
Teams need a scientific intelligence platform that unifies design, sequence, structural, assay, and biophysical attributes into a single environment, built to support workflows from antibody and protein engineering through to candidate selection.
By emphasizing a structured and scalable platform for format and panel design and multimodal integration rather than a single model, we can address the fragmentation pain-point that dominated informal discussions.
In addition, consolidating workflows for multispecific engineering, structure-aware optimization, and developability analysis can create AI-ready pipelines where models are continuously fed with real program data instead of curated one-off training sets.
This isn’t just theoretical: multimodal platforms can tangibly improve engineering of complex immunotherapies when embedded into day-to-day decision-making.
Where Dotmatics Luma fits
By supporting a more modern and cohesive enterprise architecture, Dotmatics Luma scientific intelligence platform positions organizations to accelerate and inspire innovation with the aid of automation, modeling, simulation, and AI. The category-defining platform helps teams push science forward with AI-powered data management and workflow automation for multimodal scientific discovery.
What’s unique is that with Luma, teams can automate and govern data processes across complex workflows, acquiring diverse data streams from Dotmatics and third-party instruments, equipment, software, and databases. Luma works to process, normalize, and structure data, readying it for deeper analysis using advanced modeling, simulation, and AI-based approaches that speed up and advance innovation. Luma can adapt to workflow variability and be tailored to individual preferences, providing a personalized experience for any individual user.
Recently, Dotmatics has partnered with BioGlyph to deliver AI-powered biologics-design tools from within the Dotmatics Luma platform. As part of the Dotmatics Luma Platform, BioGlyph Luma can work alongside other industry-standard software to create an end-to-end solution that optimizes both workflows and dataflows across all key steps of biologics discovery, including:
Biologic entity design and engineering
Construct design
Expression, purification and PTM analysis
Protein characterization and cytometry analysis
Where antibody engineering is headed
AE&T 2025 suggested that antibody engineering is in the middle of its digital transformation, homing in on the question: “Can we really scale the design, evaluation, and selection of complex protein formats with confidence and speed?”
The next wave will likely be shaped by four converging trends:
AI-native design loops that treat every program as a learning experiment
Data stream integration (in vitro, in vivo, and in silico discovery)
Platform-level attention to developability from the very first design step
Expansion into more diverse, non-oncology indications
If the past decade was about proving that bispecifics, ADCs, and novel protein scaffolds could work, then the next few years will be about building the infrastructure for data, models, workflows, and organizational habits to make them work reliably, repeatedly, and across indications. In that future, conferences like AE&T will evolve beyond showcasing isolated breakthroughs and toward comparing how organizations are connecting their scientific intelligence stacks where antibody engineering teams act not just as molecule designers, but as architects of the systems that design them.
Drug companies are recognizing a hard truth: they are not software companies, and pretending otherwise has cost time, money, and, most importantly, medicines.
The era of sprawling, bespoke in‑house platforms trying to be ELN, LIMS, registry, workflow engine, and analytics all at once is gone. What replaces it will determine which R&D organizations actually convert biology into drugs, and which drown in tech debt. The sooner leadership aligns their digital strategy, the sooner patients will see the benefits at the bedside.
To learn more about how Dotmatics Luma is helping transform therapeutic discovery visit: https://www.dotmatics.com/luma/antibody-and-protein-engineering.
