Back

Using machine learning to design adeno-associated virus capsids with high likelihood of viral assembly

To, C. T.; Wirsching, C.; Marques, A. D.; Zolotukhin, S.

2021-06-02 bioinformatics
10.1101/2021.05.18.444607 bioRxiv
Show abstract

We study the application of machine learning in designing adeno-associated virus (AAV) capsid sequences with high likelihood of viral assembly, i.e. capsid viability. Specifically, we design and implement Origami, a model-based optimization algorithm, to identify highly viable capsid sequences within the vast space of 2033 possibilities. Our evaluation shows that Origami performs well in terms of optimality and diversity of model-designed sequences. Moreover, these sequences are ranked according to their viability score. This helps designing experiments given budget constraint.

Matching journals

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

50% of probability mass above

"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.