Main Conference Day 1 - PT (Pacific Time, GMT-08:00)
AI/ML for biotherapeutics is constrained by the scale and quality of training data. In this session, Twist Bioscience will present multiple workflows for strategies to bridge this gap using high-fidelity synthetic DNA platforms and bespoke data outputs that integrates next-generation synthesis and production and characterization directly into the Design-Make-Test-Learn cycle. Case studies will illustrate how LLMs are validated using Twist “off-the-shelf” data sets, how high-throughput iterations of make-test cycles can be used to compare and train new models, and when in silico (de novo) designed libraries coupled with wet-lab panning and screening can simultaneously generate lead therapeutic candidates while also validating and training generative models. Join us to learn how scalable and innovative antibody services transform ML into a powerful engine for rapid biotherapeutic discovery.
- Colby Souders, PhD - Chief Scientific Officer, Biopharma, Twist Bioscience
Triaging and transitioning a large panel of prospective antibody “hits-to-leads” is de-risked when biology, developability, and manufacturability attribute assessments are integrated by design. ATUM’s services for advancing discovery sequences to a manufacturable biologic with phase appropriate assessments will be outlined. These include AI and in silico screening, high-throughput transient expression for rapid material generation and developability assessment for rank ordering hits. Selected leads are transitioned to an intermediate, higher yield system for high resolution analytics and a manufacturing predictive selection process. Finally, iterative knowledge gained is seamlessly executed in a commercial manufacturing-ready cell line.
- Jeff Johnson - Senior Scientific Director of Cell Line Development and Biologics, ATUM
Next-gen immunotherapies demand seamless integration of multimodal data—sequence, structure, assay, and biophysical insights. Traditional tools can’t keep pace. This talk introduces a new paradigm: a Multimodal Scientific Intelligence Platform built to unify antibody/protein workflows, enhance collaboration, and accelerate AI-ready discovery. Includes a case study from a major biopharma showing how multimodal workflows improve outcomes in multispecific antibody engineering.
- Christian Olsen - Vice President, Strategy - Protein Therapeutics, Luma, Dotmatics
Traditional antibody discovery approaches often prioritize single objectives, failing to balance multiple properties simultaneously which yield candidates with compromised developability profiles. We present a selection framework using Pareto optimization across rank-normalized scores with hierarchical property classification. This approach generates balanced candidate shortlists with AI-assisted explanations of property trade-offs, enabling efficient identification of optimal molecules for validation while reducing costly experimental iterations and accelerating therapeutic antibody development.
- Kemal Sonmez, PhD - Senior Applied AI Scientist, Amazon
Artificial intelligence (AI) is transforming antibody discovery and engineering. Ailux's platform synergistically combines the best of our comprehensive wet lab, AtlaX biologics database, and three proprietary AI engines. We will explore our latest case studies that exemplify our AI-driven approach for tackling challenging targets, identifying unique functional antibodies, and achieving multi-objective optimization. This presentation provides our realistic and evidence-based perspective on the impact of AI on developing next-generation antibody therapeutics.
- Barry Duplantis, PhD - Senior Director of Business Development, Ailux
Antibodies discovered in vivo have many advantages such as high affinity and low polyspecificity/reactivity with superior developability profiles compared with those identified through in vitro methods. However, immunization-driven approaches have historically faced challenges with complex targets that are evolutionarily conserved. We present a case study in which genetic immunization successfully generated a panel of antibodies against a 100% conserved, multi-subunit, cell-surface heterocomplex, and the epitope diversity was revealed through in silico mapping. This study highlights that breaking tolerance in mice can be effectively achieved and integration of in silico tools facilitate rapid decision making for downstream lead selection.
- Shuji Sato - Vice President of Innovative Solutions, MindWalk AI