Hands-on with Amazon Bio Discovery: From AI Antibody Design to Wet Lab

Amazon Bio Discovery is an AWS bio-research application that uses AI models to design and evaluate antibody candidates. It reached GA in April this year and is currently available only in the us-east-1 (Virginia) region. The platform supports a Lab-in-the-Loop environment that iterates through AI design, actual experiments, and result feedback.

Key Features and Concepts

The platform provides over 40 AI models, which can be combined to create Recipes (workflows) tailored to research goals. Designed candidates can be linked with partners such as Ginkgo Bioworks and Twist Bioscience to proceed to actual Wet Lab experiments. It also supports bringing in user-created AI models or training them with proprietary data (Beta).

Key concepts are as follows:

• Project: A workspace for managing Experiments, files, results, etc.
• Module: Individual AI models or analytical functions that make up a Recipe
• Recipe: A Workflow created by combining multiple Modules
• Experiment: The process of applying data to a Recipe, executing it, and checking results

Hands-on Session: HER2 Target Antibody Design

In the hands-on session, antibody candidates were generated targeting the HER2 protein (PDB file). An Antibody generation Recipe was selected, Design and Binding Affinity evaluation Modules were configured, and then the HER2 structure file and antibody scaffold file were uploaded to run an Experiment. Generating about 10 antibody candidates had an estimated cost of 1 EU (Experiment Unit) and a duration of 2–4 hours, with the actual execution completing in about 2 hours.

The results screen allows analysis of candidates' binding affinity, structural formation potential, etc., and verification of candidate diversity through Functional Clustering. Afterward, the Send to Web Lab feature can be used to connect to actual experiments.

Pricing and Features

Costs are charged in EU (Experiment Unit) units. New users receive 5 EU for free in the first month, and failed Experiments do not deduct EU. Wet Lab experiment costs are charged separately from EU. Currently, an Early Access discount is applied, offering a 50% discount off the regular price (until October 15, 2026).

Pros include the integrated use of diverse AI models, easy accessibility via GUI, and Wet Lab integration. On the other hand, cons include the need for initial learning of unique concepts like Project and Recipe, difficulty in checking computing resource specifications, and automatic session termination. Unlike AWS HealthOmics, which focuses on genomic data analysis, this service is specialized for protein and antibody candidate design.