PIGNet: a physics-informed deep learning model toward generalized drug-target interaction predictions.

Recently, deep neural network (DNN)-based drug-target interaction (DTI) models were highlighted for their high accuracy with affordable computational costs. Yet, the models' insufficient generalization remains a challenging problem in the practice of in silico drug discovery. We propose two key strategies to enhance generalization in the DTI model. The first is to predict the atom-atom pairwise interactions via physics-informed equations parameterized with neural networks and provides the total binding affinity of a protein-ligand complex as their sum. We further improved the model generalization by augmenting a broader range of binding poses and ligands to training data. We validated our model, PIGNet, in the comparative assessment of scoring functions (CASF) 2016, demonstrating the outperforming docking and screening powers than previous methods. Our physics-informing strategy also enables the interpretation of predicted affinities by visualizing the contribution of ligand substructures, providing insights for further ligand optimization.

Comprehensive Study on Molecular Supervised Learning with Graph Neural Networks.

This work considers strategies to develop accurate and reliable graph neural networks (GNNs) for molecular property predictions. Prediction performance of GNNs is highly sensitive to the change in various parameters due to the inherent challenges in molecular machine learning, such as a deficient amount of data samples and bias in data distribution. Comparative studies with well-designed experiments are thus important to clearly understand which GNNs are powerful for molecular supervised learning. Our work presents a number of ablation studies along with a guideline to train and utilize GNNs for both molecular regression and classification tasks. First, we validate that using both atomic and bond meta-information improves the prediction performance in the regression task. Second, we find that the graph isomorphism hypothesis proposed by [Xu, K.; et al How powerful are graph neural networks? 2018, arXiv:1810.00826. arXiv.org e-Print archive. https://arxiv.org/abs/1810.00826] is valid for the regression task. Surprisingly, however, the findings above do not hold for the classification tasks. Beyond the study on model architectures, we test various regularization methods and Bayesian learning algorithms to find the best strategy to achieve a reliable classification system. We demonstrate that regularization methods penalizing predictive entropy might not give well-calibrated probability estimation, even though they work well in other domains, and Bayesian learning methods are capable of developing reliable prediction systems. Furthermore, we argue the importance of Bayesian learning in virtual screening by showing that well-calibrated probability estimation may lead to a higher success rate.

Histogen Layers Contributing to Adventitious Bud Formation Are Determined by their Cell Division Activities.

Saintpaulia ionantha is propagated by adventitious buds in horticulture, and periclinal chimeral cultivars are usually difficult to propagate. However, some periclinal chimeral cultivars can be propagated with adventitious buds, and the mechanism of which has been unknown. Striped flower cultivars "Kaname," "Concord," and "Monique" were used to investigate what causes flower color separation in adventitious shoot-derived plants by tissue culture. These cultivars were revealed to have mutated flavonoid 3', 5' hydroxylase (SiF3'5'H), WDR1 (SiWDR1), or flavonoid 3 hydroxylase (SiF3H), respectively, in their L1 layer. From our previous study using "Kaname," all flowers from adventitious shoots were colored pink, which was the epidermal color of mother plants' flowers. We used "Concrd" and "Monique" from which we obtained not only monochromatic-colored plants the same as the epidermal color of mother plants, but also plants with a monochromatic colored plants, same as the subepidermal color, and a striped flower color the same as mother plants. Histological observations revealed that epidermal cells divided actively at 14 d after culture and they were involved in the formation of adventitious shoots in the cultured leaf segments of "Kaname." On the other hand, in "Concord" and "Monique," the number of divided cells in the subepidermis was rather higher than that of epidermal cells, and subepidermal cells were sometimes involved in shoot formation. In addition, the plant and leaf size of L1-derived plants from "Concord" and "Monique" were non-vigorous and smaller than those derived from the subepidermal layer. In conclusion, periclinal chimeral cultivars of Saintpaulia can be divided into two types. One type has a high cell division activity in the L1 layer, from which only single flower-colored plants derived from L1 can be obtained as adventitious shoots. Another type has a low cell division activity in the L1 layer, from which striped flower-colored plants the same as mother plants derived from several layers including L1 can be obtained as adventitious shoots. In the periclinal chimeral cultivar capable of propagation with adventitious shoots, the possibility was shown that cells in the L2 layer could form shoots by involving cells of the L1 layer with a low division activity.

On-Demand Drug Delivery System Using Micro-organogels with Gold Nanorods.

In this study, we designed a biocompatible drug carrier: micro-organogels prepared by emulsification using vegetable oils and self-assembled gelator fibers. Flurbiprofen was chosen as a hydrophobic model drug and is classified as a nonsteroidal anti-inflammatory drug. In the absence of NIR light, flurbiprofen encapsulated in micro-organogels with gold nanorods (GNRs) was released slowly, while release was accelerated in the presence of NIR light due to the increase in the temperature surrounding the GNRs that transforms the gels into liquid. These results suggest that our system can be efficiently used as a versatile scaffold for on-demand drug delivery systems.

Inhibition of chalcone synthase expression in anthers of Raphanus sativus with Ogura male sterile cytoplasm.

BACKGROUND AND AIMS: Expression of the mitochondrial gene orf138 causes Ogura cytoplasmic male sterility (CMS) in Raphanus sativus, but little is known about the mechanism by which CMS takes place. A preliminary microarray experiment revealed that several nuclear genes concerned with flavonoid biosynthesis were inhibited in the male-sterile phenotype. In particular, a gene for one of the key enzymes for flavonoid biosynthesis, chalcone synthase (CHS), was strongly inhibited. A few reports have suggested that the inhibition of CHS causes nuclear-dependent male sterile expression; however, there do not appear to be any reports elucidating the effect of CHS on CMS expression. In this study, the expression patterns of the early genes in the flavonoid biosynthesis pathway, including CHS, were investigated in normal and male-sterile lines. METHODS: In order to determine the aberrant stage for CMS expression, the characteristics of male-sterile anthers are observed using light and transmission electron microscopy for several stages of flower buds. The expression of CHS and the other flavonoid biosynthetic genes in the anthers were compared between normal and male-sterile types using real time RT-PCR. KEY RESULTS: Among the flavonoid biosynthetic genes analysed, the expression of CHS was strongly inhibited in the later stages of anther development in sterility cytoplasm; accumulation of putative naringenin derivatives was also inhibited. CONCLUSIONS: These results show that flavonoids play an important role in the development of functional pollen, not only in nuclear-dependent male sterility, but also in CMS.