English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Reaction prediction has been recognized as a critical task in syntheticchemistry, where the goal is to predict the outcome of a reaction based on thegiven reactants. With the widespread adoption of generative models, theVariational Autoencoder(VAE) framework has typically been employed to tacklechallenges in reaction prediction, where the reactants are encoded as acondition for the decoder, which then generates the product. Despiteeffectiveness, these conditional VAE (CVAE) models still fail to adequatelyaccount for the inherent uncertainty in reaction prediction, which primarilystems from the stochastic reaction process. The principal limitations aretwofold. Firstly, in these CVAE models, the prior is independent of thereactants, leading to a default wide and assumed uniform distribution varianceof the generated product. Secondly, reactants with analogous molecularrepresentations are presumed to undergo similar electronic transitionprocesses, thereby producing similar products. This hinders the ability tomodel diverse reaction mechanisms effectively. Since the variance in outcomesis inherently non-uniform, we are thus motivated to develop a framework thatgenerates reaction products with non-uniform uncertainty. Firstly, we eliminatethe latent variable in previous CVAE models to mitigate uncontrol-label noise.Instead, we introduce randomness into product generation via boosting toensemble diverse models and cover the range of potential outcomes, and throughdropout to secure models with minor variations. Additionally, we design aranking method to union the predictions from boosting and dropout, prioritizingthe most plausible products. Experimental results on the largest reactionprediction benchmark USPTO-MIT show the superior performance of our proposedmethod in modeling the non-uniform uncertainty compared to baselines.

Modeling non-uniform uncertainty in Reaction Prediction via Boosting and Dropout