Korro Bio: Development and Application of Graph Neural Networks to Design Guide Oligonucleotides that Promote RNA Editing

U n l o c k i n g t h e O P E R A P l a t f o r m

Development and Application of Graph

Neural Networks to Design Guide

Oligonucleotides that Promote RNA Editing

Overview of Topics

1. Korro Bio Methodology
2. Machine Learning for Chemically Modified Oligonucleotide Design
3. Methods for Increasing Use-Cases for ML Oligonucleotide Design
4. • New Chemical Modifications
   • New Targets

2

RNA Editing: Transiently Effecting anA-to-I Edit on RNA Using anOligonucleotide

1

2 Oligo-RNA duplex recruits adenosine deaminase acting on RNA (ADAR)

3 ADAR catalyzes deamination: 'A' to 'I' edit

DNA with disease-	ADAR		ADAR
	Adenosine	Inosine
causing mutation	Target
	RNA

4 mRNA translated to protein with 'I' read as 'G'

5 Resultant therapeutic protein

3

OPERA: Our Differentiated Approach for RNA Editing

Deep understanding of	Expertise in oligonucleotide
ADAR biology	chemistry

Fit-for-purpose delivery	Machine learning optimization
of oligonucleotides

1 IP estate count as of September 18, 2023	4

Chemical Modifications are Employed to Drive Activity and Stability

Examples of chemical modifications in

approved products

Korro leverages established and novel chemical modifications to improve pharmacology of ADAR guide oligonucleotides

5

Chemical Modification Pattern Significantly Impacts Editing of Target mRNA

Editing

Sequence

6

Overview of Topics

1. Korro Bio Methodology
2. Machine Learning for Chemically Modified Oligonucleotide Design
3. Methods for Increasing Use-Cases for ML Oligonucleotide Design
4. • New Chemical Modifications
   • New Targets

7

Sequence-Level Featurization

Structural Features

Entity Embeddings

8

Sequence-Level Featurization

Tree-based Models

Entity Embeddings

Convolutional Neural Network

9

Our Model Learns the Effect of Chemical Modification Patterns on In Vitro Editing

r = 0.80

• Model tested on 20% of in vitro data for a single target

• Understanding of relationship between sequence, chemical modifications and in vitro editing

• Predictions within 7% of in vitro editing

10