NeoBody™: Lab-in-the-loop AI for Highest Possible Affinities
Pre-trained Ab language model (> 10 )

10
High-throughput screen by iHTS™
.jpg)
Reference Ab seq
> 10,000 variants with diversity amongst all variants
Automated Affinity scoring and ranking model
.jpg)
Most comprehensive optimized novel mabs with industry leading affinities
Bayesian-based exhaustive in silico screen
.jpg)
AbMap™: Proprietary AI antibody sequences allowing maximum diversity and affinity gains
Multi-granularity antibody database
Abs with known Ags
Abs from human
Cancer specific
Al-generated VHHs
Advantages of the proprietary database
Greatest Diversity
(Proprietary database generated internally)
Increasing Rate
(> 100%/Y)
Highest Quality
(H-L paired;Ag-Ab paired)
Database Expanding
(> 10^10)
Case Study: Best-in-Class Clinical Stage Ab Created
Challenges
-
Maximize affinity and generate novel IP while maintaining all functionality;
-
Must bind to identical epitope of the given antibody;
-
Affinity to be orders of magnitude higher than given antibody;
-
Internalization rate had to be similar to the given antibody;
-
Program to be delivered in 2 months.
1st Deep-Learning Modeling Run
-
Generate 100K sequences from > 10 sequences.
-
Construct a phage library and high-throughput screen by iHTS™
10
Deep-Learning
-
Generate 20 sequences.
-
Affinity, internalization, and epitope test.
Results
-
Best-in-Class mab created with FTO and new IP generated;
-
Superior clinical properties to incumbent mabs in clinic.
-
Specificity to epitope achieved;
-
Orders of magniture higher affinity than the reference Ab.
.jpg)
Affinity Dramatically Increased Against Same Epitope
.jpg)
Consistent internalization acheived
.jpg)