Back

DyAb: sequence-based antibody design and property prediction in a low-data regime

Lin, J. Y.-Y.; Hofmann, J. L.; Leaver-Fay, A.; Liang, W.-C.; Vasilaki, S.; Lee, E.; Pinheiro, P. O.; Tagasovska, N.; Kiefer, J. R.; Wu, Y.; Seeger, F.; Bonneau, R.; Gligorijevic, V.; Watkins, A.; Cho, K.; Frey, N. C.

2025-02-02 bioengineering Community evaluation
10.1101/2025.01.28.635353 bioRxiv
Show abstract

Protein therapeutic design and property prediction are frequently hampered by data scarcity. Here we propose a new model, DyAb, that addresses these issues by leveraging a pair-wise representation to predict differences in protein properties, rather than absolute values. DyAb is built on top of a pre-trained protein language model and achieves a Spearman rank correlation of up to 0.85 on binding affinity prediction across molecules targeting three different antigens (EGFR, IL-6, and an internal target), given as few as 100 training data. We employ DyAb in two design contexts: as a ranking model to score combinations of known mutations, and combined with a genetic algorithm to generate new sequences. Our method consistently generates novel antibody candidates with high binding rates, including designs that improve on the binding affinity of the lead molecule by more than ten-fold. DyAb represents a powerful tool for engineering therapeutic protein properties in low data regimes common in early-stage drug development.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

1
Nature Communications
○ 5641 papers in training set
Top 5%
25.9%
Similar papers in this journal
2
Nature Methods
○ 385 papers in training set
Top 1%
9.5%
Similar papers in this journal
3
Proceedings of the National Academy of Sciences
● 2444 papers in training set
Top 7%
6.1%
Similar papers in this journal
4
Nature Machine Intelligence
○ 70 papers in training set
Top 0.5%
5.3%
Similar papers in this journal
5
Science
● 477 papers in training set
Top 2%
4.7%
Similar papers in this journal
50% of probability mass above
6
Cell Systems
○ 201 papers in training set
Top 0.9%
4.7%
7
Nature Biomedical Engineering
○ 47 papers in training set
Top 0.3%
4.2%
8
Nature Biotechnology
○ 172 papers in training set
Top 1%
3.9%
9
Science Advances
● 1243 papers in training set
Top 12%
3.1%
10
Nature
○ 645 papers in training set
Top 5%
2.6%
11
ACS Central Science
● 71 papers in training set
Top 0.5%
2.4%
12
Cell Chemical Biology
○ 94 papers in training set
Top 0.6%
2.1%
13
Communications Biology
○ 993 papers in training set
Top 11%
2.1%
14
eLife
● 5828 papers in training set
Top 54%
1.5%
15
Cell Reports
○ 1498 papers in training set
Top 22%
1.3%
16
Nature Chemical Biology
○ 119 papers in training set
Top 2%
1.3%
17
Cell Genomics
○ 172 papers in training set
Top 3%
1.3%
18
Journal of the American Chemical Society
● 217 papers in training set
Top 2%
1.1%
19
mAbs
○ 32 papers in training set
Top 0.3%
1.1%
20
Nature Chemistry
○ 42 papers in training set
Top 0.7%
1.1%
21
Nature Computational Science
○ 55 papers in training set
Top 1%
1.0%
22
Angewandte Chemie International Edition
○ 93 papers in training set
Top 2%
0.9%
23
Neuron
○ 337 papers in training set
Top 5%
0.8%
24
Scientific Reports
○ 3612 papers in training set
Top 74%
0.8%
25
Advanced Science
○ 286 papers in training set
Top 12%
0.6%

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.