Decomposition Profile Data Analysis for Deep Understanding of Multiple Effects of Natural Products

Nemoto, Shumpei; Morita, Katsuaki; Mizuno, Tadahaya; Kusuhara, Hiroyuki

doi:10.1021/acs.jnatprod.0c01381

Cited by 14 publications

(7 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The numerical representation of chemicals is useful for grasping their properties and is roughly divided into two approaches: phenotype-based and structure-based representation [1][2][3]. The latter is often called a descriptor, and determining how to generate a good descriptor is one of the big topics in the field of chemoinformatics [4,5].…”

Section: Introductionmentioning

confidence: 99%

Investigation of chemical structure recognition by encoder–decoder models in learning progress

2023

Self Cite

View full text Add to dashboard Cite

Descriptor generation methods using latent representations of encoder–decoder (ED) models with SMILES as input are useful because of the continuity of descriptor and restorability to the structure. However, it is not clear how the structure is recognized in the learning progress of ED models. In this work, we created ED models of various learning progress and investigated the relationship between structural information and learning progress. We showed that compound substructures were learned early in ED models by monitoring the accuracy of downstream tasks and input–output substructure similarity using substructure-based descriptors, which suggests that existing evaluation methods based on the accuracy of downstream tasks may not be sensitive enough to evaluate the performance of ED models with SMILES as descriptor generation methods. On the other hand, we showed that structure restoration was time-consuming, and in particular, insufficient learning led to the estimation of a larger structure than the actual one. It can be inferred that determining the endpoint of the structure is a difficult task for the model. To our knowledge, this is the first study to link the learning progress of SMILES by ED model to chemical structures for a wide range of chemicals. Graphical Abstract

show abstract

Section: Introductionmentioning

confidence: 99%

Investigation of chemical structure recognition by encoder–decoder models in learning progress

2023

Self Cite

View full text Add to dashboard Cite

show abstract

“…This is because compound administration is relatively easy to control and less confounded compared with other model preparation approaches such as diet and surgery, making it easier to compare findings. In addition, considering that compounds generally have multiple effects, including those unrecognized, it is also expected that a diverse spectrum of biological responses can be covered due to nonarbitrary perturbations by various compounds (Morita et al ., 2020; Nemoto et al ., 2021). Open TG-GATEs, for example, contain transcriptome data, pathological images, and blood biochemistry of tissues collected at various time points from rats treated with various compounds at various concentrations, while there is no information regarding immune cells in tissue in these databases.…”

Section: Introductionmentioning

confidence: 99%

Rat Deconvolution as Knowledge Miner for Immune Cell Trafficking from Toxicogenomics Databases

Morita

Mizuno

Azuma

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Toxicogenomics databases are useful for understanding biological responses in individuals because they are derived from well-controlled experiments and include a diverse spectrum of biological responses. Although these database contain no information regarding immune cells in the liver, which are important in the progression of liver injury, deconvolution that estimates cell-type proportions from bulk transcriptome could add information regarding immune cell trafficking to the database. However, deconvolution has been mainly applied to humans and mice and less often to rats, which are the main target of toxicogenomics databases. Here, we developed a deconvolution method for rats and established a methodology to obtain information regarding immune cells from toxicogenomics databases. The contributions of this work are three-fold. First, we obtained the gene expression profiles of various rat immune cells necessary for deconvolution and constructed a dataset; second, we compared the accuracy of models based on human and mouse datasets and showed the impact of species differences on deconvolution; third, we showed that rat deconvolution could retrieve information regarding immune cell trafficking from toxicogenomics databases.

show abstract

“…These effects can be described as the biological responses of an organism to the compound [3], [4]. Numerization and subsequent data-driven analysis are valuable in understanding compound effects, including toxicity prediction [5], [6]. Pathological imaging, specifically toxicopathology, plays a crucial role in safety assessment of compounds, providing a repository of in vivo biological responses from experimental animals exposed to specific compounds [7].…”

Section: Introductionmentioning

confidence: 99%

Investigation of latent representation of toxicopathological images extracted by CNN model for understanding compound propertiesin vivo

Maedera

Mizuno

Kusuhara

2023

Preprint

Self Cite

View full text Add to dashboard Cite

Toxicopathological images acquired during safety assessment elucidate an individual's biological responses to a given compound, and their numerization can yield valuable insights contributing to the assessment of compound properties. Currently, toxicopathological images are mainly encoded as pathological findings, evaluated by pathologists, which introduces challenges when used as input for modeling, specifically in terms of representation capability and comparability. In this study, we assessed the usefulness of latent representations extracted from toxicopathological images using Convolutional Neural Network (CNN) in estimating compound properties in vivo. Special emphasis was placed on examining the impact of learning pathological findings, the depth of frozen layers during learning, and the selection of the layer for latent representation. Our findings demonstrate that a machine learning model fed with the latent representation as input surpassed the performance of a model directly employing pathological findings as input, particularly in the classification of a compound's Mechanism of Action and in predicting late-phase findings from early-phase images in repeated-dose tests. While learning pathological findings did improve accuracy, the magnitude of improvement was relatively modest. Similarly, the effect of freezing layers during learning was also limited. Notably, the selection of the layer for latent representation had a substantial impact on the accurate estimation of compound properties in vivo.

show abstract

Decomposition Profile Data Analysis for Deep Understanding of Multiple Effects of Natural Products

Cited by 14 publications

References 52 publications

Investigation of chemical structure recognition by encoder–decoder models in learning progress

Investigation of chemical structure recognition by encoder–decoder models in learning progress

Rat Deconvolution as Knowledge Miner for Immune Cell Trafficking from Toxicogenomics Databases

Investigation of latent representation of toxicopathological images extracted by CNN model for understanding compound propertiesin vivo

Contact Info

Product

Resources

About