Simplified, interpretable graph convolutional neural networks for small molecule activity prediction

We here present a streamlined, explainable graph convolutional neural network (gCNN) architecture for small molecule activity prediction. We first conduct a hyperparameter optimization across nearly 800 protein targets that produces a simplified gCNN QSAR architecture, and we observe that such a model can yield performance improvements over both standard gCNN and RF methods on difficult-to-classify test sets. Additionally, we discuss how reductions in convolutional layer dimensions potentially speak to the “anatomical” needs of gCNNs with respect to radial coarse graining of molecular substructure. We augment this simplified architecture with saliency map technology that highlights molecular substructures relevant to activity, and we perform saliency analysis on nearly 100 data-rich protein targets. We show that resultant substructural clusters are useful visualization tools for understanding substructure-activity relationships. We go on to highlight connections between our models’ saliency predictions and observations made in the medicinal chemistry literature, focusing on four case studies of past lead finding and lead optimization campaigns.

Inaction on Lead Despite the Relevant Knowledge: Predictors, Covariates, and Outreach Implications

Testing residential soil and paint for lead provides actionable information. By showing where and how much lead exists on the residence, it allows one to quantify risk and determine the best ways to reduce exposure along with the corresponding health and financial costs. For these reasons, several federal and state programs offer outreach to audiences on the benefits of testing residential soil and paint for lead. Not all individuals who know about lead’s adverse health effects, however, test their residence for lead, potentially limiting the actionable information that could have helped to reduce their exposure. Such individuals represent a challenge to outreach programs and the broader public health objectives. There is, thus, a need to understand who such individuals are and why they behave this way, allowing us to develop a specialized outreach program that addresses the problem by targeting the relevant sub-population. Using survey data, we quantitatively determine the profiles of individuals who, despite knowing about lead’s adverse health effects, are unlikely to test their residence for lead, finding statistically significant socio-economic predictors and behavioral covariates. We also find a geographic component to it, further helping outreach professionals learn how to allocate their limited resources.

Encoded Library Technologies as Integrated Lead Finding Platforms for Drug Discovery

The scope of targets investigated in pharmaceutical research is continuously moving into uncharted territory. Consequently, finding suitable chemical matter with current compound collections is proving increasingly difficult. Encoded library technologies enable the rapid exploration of large chemical space for the identification of ligands for such targets. These binders facilitate drug discovery projects both as tools for target validation, structural elucidation and assay development as well as starting points for medicinal chemistry. Novartis internalized two complementing encoded library platforms to accelerate the initiation of its drug discovery programs. For the identification of low-molecular weight ligands, we apply DNA-encoded libraries. In addition, encoded peptide libraries are employed to identify cyclic peptides. This review discusses how we apply these two platforms in our research and why we consider it beneficial to run both pipelines in-house.

Efficient Lead Finding, Activity Enhancement and Preliminary Selectivity Control of Nuclear Receptor Ligands Bearing a Phenanthridinone Skeleton

Background: Nuclear receptors (NRs) are considered as potential drug targets because they control diverse biological functions. However, steroidal ligands for NRs have the potential to cross-react with other nuclear receptors, so development of non-steroidal NR ligands is desirable to obtain safer agents for clinical use. We anticipated that efficient lead finding and enhancement of activity toward nuclear receptors recognizing endogenous steroidal ligands might be achieved by exhaustive evaluation of a steroid surrogate library coupled with examination of structure-activity relationships (SAR). Method: We evaluated our library of RORs (retinoic acid receptor-related orphan receptors) inverse agonists and/or PR (progesterone receptor) antagonists based on the phenanthridinone skeleton for antagonistic activities toward liver X receptors (LXRs), androgen receptor (AR) and glucocorticoid receptor (GR) and examined their SAR. Results: Potent LXRβ, AR, and GR antagonists were identified. SAR studies led to a potent AR antagonist (IC50: 0.059 μM). Conclusions: Our approach proved effective for efficient lead finding, activity enhancement and preliminary control of selectivity over other receptors. The phenanthridinone skeleton appears to be a promising steroid surrogate.

METAANALISIS KORELASI ANTARA MOTIVASI DENGAN ADOPSI INOVASI PERTANIAN

Some researches used the same variables as independent or dependent variables often have different conclusion. For example, a research has a conclusion that attitude toward new technology influence significantly the adoption of the technology, but another researcher has conclusion that attitude does not influence the adoption, etc. Metaanalysis is useful to understand why results of the researches are imperfect, and moreover one maybe find new variables that can lead finding new theory. This analysis want to try to understand problem on correlation between motivation and adoption of agriculture inovation from ten thesis S2 and SI.