GraphXCOVID: Explainable deep graph diffusion pseudo-Labelling for identifying COVID-19 on chest X-rays.

Can one learn to diagnose COVID-19 under extreme minimal supervision? Since the outbreak of the novel COVID-19 there has been a rush for developing automatic techniques for expert-level disease identification on Chest X-ray data. In particular, the use of deep supervised learning has become the go-to paradigm. However, the performance of such models is heavily dependent on the availability of a large and representative labelled dataset. The creation of which is a heavily expensive and time consuming task, and especially imposes a great challenge for a novel disease. Semi-supervised learning has shown the ability to match the incredible performance of supervised models whilst requiring a small fraction of the labelled examples. This makes the semi supervised paradigm an attractive option for identifying COVID-19. In this work, we introduce a graph based deep semi-supervised framework for classifying COVID-19 from chest X-rays. Our framework introduces an optimisation model for graph diffusion that reinforces the natural relation among the tiny labelled set and the vast unlabelled data. We then connect the diffusion prediction output as pseudo-labels that are used in an iterative scheme in a deep net. We demonstrate, through our experiments, that our model is able to outperform the current leading supervised model with a tiny fraction of the labelled examples. Finally, we provide attention maps to accommodate the radiologist's mental model, better fitting their perceptual and cognitive abilities. These visualisation aims to assist the radiologist in judging whether the diagnostic is correct or not, and in consequence to accelerate the decision.

Axially Chiral Enamides: Substituent Effects, Rotation Barriers, and Implications for their Cyclization Reactions.

The barrier to rotation around the N-alkenyl bond of 38 N-alkenyl-N-alkylacetamide derivatives was measured (ΔG(⧧) rotation varied between <8.0 and 31.0 kcal mol(-1)). The most important factor in controlling the rate of rotation was the level of alkene substitution, followed by the size of the nitrogen substituent and, finally, the size of the acyl substituent. Tertiary enamides with four alkenyl substituents exhibited half-lives for rotation between 5.5 days and 99 years at 298 K, sufficient to isolate enantiomerically enriched atropisomers. The radical cyclizations of a subset of N-alkenyl-N-benzyl-α-haloacetamides exhibiting relatively high barriers to rotation round the N-alkenyl bond (ΔG(⧧) rotation >20 kcal mol(-1)) were studied to determine the regiochemistry of cyclization. Those with high barriers (>27 kcal mol(-1)) did not lead to cyclization, but those with lower values produced highly functionalized γ-lactams via a 5-endo-trig radical-polar crossover process that was terminated by reduction, an unusual cyclopropanation sequence, or trapping with H2O, depending upon the reaction conditions. Because elevated temperatures were necessary for cyclization, this precluded study of the asymmetric transfer in the reaction of individual atropisomers. However, enantiomerically enriched atropsiomeric enamides should be regarded as potential asymmetric building blocks for reactions that can be accomplished at room temperature.

Synthesis of 2,6-trans- and 3,3,6-trisubstituted tetrahydropyran-4-ones from Maitland-Japp derived 2H-dihydropyran-4-ones: a total synthesis of diospongin B.

6-Substituted-2H-dihydropyran-4-one products of the Maitland-Japp reaction have been converted into tetrahydropyrans containing uncommon substitution patterns. Treatment of 6-substituted-2H-dihydropyran-4-ones with carbon nucleophiles led to the formation of tetrahydropyran rings with the 2,6-trans-stereochemical arrangement. Reaction of the same 6-substituted-2H-dihydropyran-4-ones with l-Selectride led to the formation of 3,6-disubstituted tetrahydropyran rings, while trapping of the intermediate enolate with carbon electrophiles in turn led to the formation 3,3,6-trisubstituted tetrahydropyran rings. The relative stereochemical configuration of the new substituents was controlled by the stereoelectronic preference for pseudo-axial addition of the nucleophile and trapping of the enolate from the opposite face. Application of these methods led to a synthesis of the potent anti-osteoporotic diarylheptanoid natural product diospongin B.

A Maitland-Japp inspired synthesis of dihydropyran-4-ones and their stereoselective conversion to functionalised tetrahydropyran-4-ones.

The Maitland-Japp reaction has been extended to the synthesis of highly functionalised dihydropyran-4-ones. These dihydropyran-4-ones can in turn be converted stereoselectively into tetrahydropyran-4-ones with tertiary and quaternary stereocentres via the one-pot addition of hydride or carbon nucleophiles and trapping with carbon electrophiles. The utility of this method is demonstrated by providing access to the functionalised tetrahydropyran units present in a component of the Civet fragrance and the anticancer polyketide lasonolide A.

LaplaceNet: A Hybrid Graph-Energy Neural Network for Deep Semisupervised Classification.

Semisupervised learning (SSL) has received a lot of recent attention as it alleviates the need for large amounts of labeled data which can often be expensive, requires expert knowledge, and be time consuming to collect. Recent developments in deep semisupervised classification have reached unprecedented performance and the gap between supervised and SSL is ever-decreasing. This improvement in performance has been based on the inclusion of numerous technical tricks, strong augmentation techniques, and costly optimization schemes with multiterm loss functions. We propose a new framework, LaplaceNet, for deep semisupervised classification that has a greatly reduced model complexity. We utilize a hybrid approach where pseudolabels are produced by minimizing the Laplacian energy on a graph. These pseudolabels are then used to iteratively train a neural-network backbone. Our model outperforms state-of-the-art methods for deep semisupervised classification, over several benchmark datasets. Furthermore, we consider the application of strong augmentations to neural networks theoretically and justify the use of a multisampling approach for SSL. We demonstrate, through rigorous experimentation, that a multisampling augmentation approach improves generalization and reduces the sensitivity of the network to augmentation.