JAEGER – Hunting for Antimalarials with Generative Chemistry

20 July 2021, Version 1
This content is a preprint and has not undergone peer review at the time of posting.


Recent advances in generative modeling allow designing novel compounds through deep neural networks. One such neural network model, the Junction Tree Variational Auto- Encoder (JT-VAE), excels at proposing chemically valid structures. Based on JT-VAE, we built a generative modeling approach (JAEGER) for finding novel chemical matter with desired bioactivity. Using JAEGER, we designed compounds to inhibit malaria. To prioritize the compounds for synthesis, we used the in-house Profile-QSAR (pQSAR) program, a massively-multitask bioactivity model based on 12,000 Novartis assays. Based on the pQSAR activity predictions, we selected, synthesized, and experimentally profiled two compounds. Both compounds exhibited low nanomolar activity in a malaria proliferation assay as well as a biochemical assay measuring activity against PI(4)K, which is an essential kinase that regulates intracellular development in malaria. The compounds also showed low activity in a cytotoxicity assay. Our findings show that JAEGER is a viable approach for finding novel active compounds for drug discovery.


generative chemistry
junction tree variational autoencoder
machine learning

Supplementary materials

Supplementary Material
Supplementary Figures and Supplementary Note


Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.