Personalized and Explainable Employee Training Course Recommendations: A Bayesian Variational Approach

As a major component of strategic talent management, learning and development (L&D) aims at improving the individual and organization performances through planning tailored training for employees to increase and improve their skills and knowledge. While many companies have developed the learning management systems (LMSs) for facilitating the online training of employees, a long-standing important issue is how to achieve personalized training recommendations with the consideration of their needs for future career development. To this end, in this article, we present a focused study on the explainable personalized online course recommender system for enhancing employee training and development. Specifically, we first propose a novel end-to-end hierarchical framework, namely Demand-aware Collaborative Bayesian Variational Network (DCBVN), to jointly model both the employees’ current competencies and their career development preferences in an explainable way. In DCBVN, we first extract the latent interpretable representations of the employees’ competencies from their skill profiles with autoencoding variational inference based topic modeling. Then, we develop an effective demand recognition mechanism for learning the personal demands of career development for employees. In particular, all the above processes are integrated into a unified Bayesian inference view for obtaining both accurate and explainable recommendations. Furthermore, for handling the employees with sparse or missing skill profiles, we develop an improved version of DCBVN, called the Demand-aware Collaborative Competency Attentive Network (DCCAN) framework , by considering the connectivity among employees. In DCCAN, we first build two employee competency graphs from learning and working aspects. Then, we design a graph-attentive network and a multi-head integration mechanism to infer one’s competency information from her neighborhood employees. Finally, we can generate explainable recommendation results based on the competency representations. Extensive experimental results on real-world data clearly demonstrate the effectiveness and the interpretability of both of our frameworks, as well as their robustness on sparse and cold-start scenarios.

Crystal structure and functional implication of bacterial STING

Mammalian innate immune sensor STING (STimulator of INterferon Gene) was recently found to originate from bacteria. During phage infection, bacterial STING sense c-di-GMP generated by the CD-NTase (cGAS/DncV-like nucleotidyltransferase) encoded in the same operon and signal suicide commitment as a defense strategy that restricts phage propagation. However, the precise binding mode of c-di-GMP to bacterial STING and the specific recognition mechanism are still elusive. Here, we determine two complex crystal structures of bacterial STING/c-di-GMP, which provide a clear picture of how c-di-GMP is distinguished from other cyclic dinucleotides. The protein-protein interactions further reveal the driving force behind filament formation of bacterial STING. Finally, we group the bacterial STING into two classes based on the conserved motif in β-strand lid, which dictate their ligand specificity and oligomerization mechanism, and propose an evolution-based model that describes the transition from c-di-GMP-dependent signaling in bacteria to 2’3’-cGAMP-dependent signaling in eukaryotes.

Structural insights into the ion selectivity of the MgtE channel for Mg2+ over Ca2+

MgtE is a Mg2+-selective ion channel whose orthologs are widely distributed from prokaryotes to eukaryotes, including humans, and play an important role in the maintenance of cellular Mg2+ homeostasis. Previous functional analyses showed that MgtE transports divalent cations with high selectivity for Mg2+ over Ca2+. Whereas the high-resolution structure determination of the MgtE transmembrane (TM) domain in complex with Mg2+ ions revealed a Mg2+ recognition mechanism of MgtE, the previous Ca2+-bound structure of the MgtE TM domain was determined only at moderate resolution (3.2 angstrom resolution), which was insufficient to visualize the water molecules coordinated to Ca2+ ions. Thus, the structural basis of the ion selectivity of MgtE for Mg2+ over Ca2+ has remained unclear. Here, we showed that the metal-binding site of the MgtE TM domain binds to Mg2+ ~500-fold more strongly than Ca2+. We then determined the crystal structure of the MgtE TM domain in complex with Ca2+ ions at a higher resolution (2.5 angstrom resolution), allowing us to reveal hexahydrated Ca2+, which is similarly observed in the previously determined Mg2+-bound structure but with extended metal-oxygen bond lengths. Our structural, biochemical, and computational analyses provide mechanistic insights into the ion selectivity of MgtE for Mg2+ over Ca2+.

DFT-D4 Insight into the Inclusion of Amphetamine and Methamphetamine in Cucurbit[7]uril: Energetic, Structural and Biosensing Properties

The host–guest interactions of cucurbit[7]uril (CB[7]) as host and amphetamine (AMP), methamphetamine (MET) and their enantiomeric forms (S-form and R-form) as guests were computationally investigated using density functional theory calculations with the recent D4 atomic-charge dependent dispersion corrections. The analysis of energetic, structural and electronic properties with the aid of frontier molecular orbital analysis, charge decomposition analysis (CDA), extended charge decomposition analysis (ECDA) and independent gradient model (IGM) approach allowed to characterize the host–guest interactions in the studied systems. Energetic results indicate the formation of stable non-covalent complexes where R-AMP@CB[7] and S-AMP@CB[7] are more stable thermodynamically than R-MET@CB[7] and S-MET@CB[7] in gas phase while the reverse is true in water solvent. Based on structural analysis, a recognition mechanism is proposed, which suggests that the synergistic effect of van der Waals forces, ion–dipole interactions, intermolecular charge transfer interactions and intermolecular hydrogen bonding is responsible for the stabilization of the complexes. The geometries of the complexes obtained theoretically are in good agreement with the X-ray experimental structures and indicate that the phenyl ring of amphetamine and methamphetamine is deeply buried into the cavity of CB[7] through hydrophobic interactions while the ammonium group remains outside the cavity to establish hydrogen bonds with the portal oxygen atoms of CB[7].

Crystal structure of human METTL6, the m3C methyltransferase

In tRNA, the epigenetic m3C modification at position 32 in the anticodon loop is highly conserved in eukaryotes, which maintains the folding and basepairing functions of the anticodon. However, the responsible enzymes METTL2 and METTL6 were identified only in recent years. The loss of human METTL6 (hMETTL6) affects the translational process and proteostasis in cells, while in mESCs cells, it leads to defective pluripotency potential. Despite its important functions, the catalytic mechanism of the C32 methylation by this enzyme is poorly understood. Here we present the 1.9 Å high-resolution crystal structure of hMETTL6 bound by SAH. The key residues interacting with the ligand were identified and their roles were confirmed by ITC. We generated a docking model for the hMETTL6-SAH-CMP ternary complex. Interestingly, the CMP molecule binds into a cavity in a positive patch with the base ring pointing to the inside, suggesting a flipped-base mechanism for methylation. We further generated a model for the quaternary complex with tRNASer as a component, which reasonably explained the biochemical behaviors of hMETTL6. Taken together, our crystallographic and biochemical studies provide important insight into the molecular recognition mechanism by METTL6 and may aid in the METTL-based rational drug design in the future.

Recognition of the antigen-presenting molecule MR1 by a Vδ3+ γδ T cell receptor

Unlike conventional αβ T cells, γδ T cells typically recognize nonpeptide ligands independently of major histocompatibility complex (MHC) restriction. Accordingly, the γδ T cell receptor (TCR) can potentially recognize a wide array of ligands; however, few ligands have been described to date. While there is a growing appreciation of the molecular bases underpinning variable (V)δ1+ and Vδ2+ γδ TCR-mediated ligand recognition, the mode of Vδ3+ TCR ligand engagement is unknown. MHC class I–related protein, MR1, presents vitamin B metabolites to αβ T cells known as mucosal-associated invariant T cells, diverse MR1-restricted T cells, and a subset of human γδ T cells. Here, we identify Vδ1/2− γδ T cells in the blood and duodenal biopsy specimens of children that showed metabolite-independent binding of MR1 tetramers. Characterization of one Vδ3Vγ8 TCR clone showed MR1 reactivity was independent of the presented antigen. Determination of two Vδ3Vγ8 TCR-MR1-antigen complex structures revealed a recognition mechanism by the Vδ3 TCR chain that mediated specific contacts to the side of the MR1 antigen-binding groove, representing a previously uncharacterized MR1 docking topology. The binding of the Vδ3+ TCR to MR1 did not involve contacts with the presented antigen, providing a basis for understanding its inherent MR1 autoreactivity. We provide molecular insight into antigen-independent recognition of MR1 by a Vδ3+ γδ TCR that strengthens an emerging paradigm of antibody-like ligand engagement by γδ TCRs.