A Bayesian Network Approach to Explainable Reinforcement Learning with Distal Information.

Nowadays, Artificial Intelligence systems have expanded their competence field from research to industry and daily life, so understanding how they make decisions is becoming fundamental to reducing the lack of trust between users and machines and increasing the transparency of the model. This paper aims to automate the generation of explanations for model-free Reinforcement Learning algorithms by answering "why" and "why not" questions. To this end, we use Bayesian Networks in combination with the NOTEARS algorithm for automatic structure learning. This approach complements an existing framework very well and demonstrates thus a step towards generating explanations with as little user input as possible. This approach is computationally evaluated in three benchmarks using different Reinforcement Learning methods to highlight that it is independent of the type of model used and the explanations are then rated through a human study. The results obtained are compared to other baseline explanation models to underline the satisfying performance of the framework presented in terms of increasing the understanding, transparency and trust in the action chosen by the agent.

Notch Ligand Delta-Like 1 Is Associated With Loss of Vascular Endothelial Barrier Function.

Vascular leakage associated with vascular endothelial cell (vEC) dysfunction is a hallmark of sepsis. Causative for the decreased integrity of the vascular endothelium (vE) is a complex concurrence of pathogen components, inflammation-associated host factors, and the interaction of vECs and activated circulating immune cells. One signaling pathway that regulates the integrity of the vE is the Notch cascade, which is activated through the binding of a Notch ligand to its respective Notch receptor. Recently, we showed that the soluble form of the Notch ligand Delta-like1 (sDLL1) is highly abundant in the blood of patients with sepsis. However, a direct connection between DLL1-activated Notch signaling and loss of vEC barrier function has not been addressed so far. To study the impact of infection-associated sDLL1, we used human umbilical vein cells (HUVEC) grown in a transwell system and cocultured with blood. Stimulation with sDLL1 induced activation as well as loss of endothelial tight structure and barrier function. Moreover, LPS-stimulated HUVEC activation and increase in endothelial cell permeability could be significantly decreased by blocking DLL1-receptor binding and Notch signaling, confirming the involvement of the cascade in LPS-mediated endothelial dysfunction. In conclusion, our results suggest that during bacterial infection and LPS recognition, DLL1-activated Notch signaling is associated with vascular permeability. This finding might be of clinical relevance in terms of preventing vascular leakage and the severity of sepsis.

Phenotypic Detection of Hemin-Inducible Trimethoprim-Sulfamethoxazole Heteroresistance in Staphylococcus aureus.

Trimethoprim-sulfamethoxazole (SXT) is a valuable second-line antimicrobial agent to treat methicillin-resistant Staphylococcus aureus infections. Discrepancies between various antibiotic susceptibility testing (AST) methods for SXT susceptibility in S. aureus have been described. Here, we describe a hemin-inducible heteroresistance phenotype in S. aureus. We compared the results of the Vitek 2 AST on a set of 95 S. aureus clinical isolates with broth microdilution, disk diffusion using standard Mueller-Hinton agar, and disk diffusion using Mueller-Hinton agar supplemented with 5% horse blood (MHF). To investigate the potential clinical relevance of SXT heteroresistance, an in vivo Galleria mellonella infection assay was performed. All Vitek 2 SXT-susceptible (n = 17) isolates were concordant with AST results by other methods applied in this study. In 32/78 (41%) of Vitek 2 SXT-resistant isolates, we observed a heteroresistant growth phenotype on MHF. The heteroresistance phenotype was associated with the presence of dfr genes, encoding trimethoprim resistance. The addition of a hemin-impregnated disk in a double disk diffusion method on standard Mueller-Hinton agar was able to induce growth in the SXT zone of inhibition. An in vivo infection assay with G. mellonella suggested that the SXT heteroresistance phenotype resulted in lethality similar to that of the SXT-resistant phenotype. In this study, we describe a novel hemin-inducible heteroresistance phenotype in S. aureus. This heteroresistance phenotype may be missed by standard AST methods but can be detected by performing disk diffusion using Mueller-Hinton agar supplemented with 5% horse blood, commonly used for AST of fastidious organisms. This phenomenon may partly explain the discrepancies of AST methods in determining SXT resistance in S. aureus. IMPORTANCE Staphylococcus aureus is one of most important pathogens in clinical medicine. Besides its virulence, the acquisition or emergence of resistance toward antibiotic agents, in particular to beta-lactam antibiotics (methicillin-resistant S. aureus [MRSA]), poses a major therapeutic challenge. Trimethoprim-sulfamethoxazole (SXT) is one of the effective antimicrobial agents of last resort to treat MRSA infections. Here, we report the detection of a SXT-heteroresistant phenotype which is inducible by hemin and can be detected using Mueller-Hinton agar supplemented with horse blood. Heteroresistance describes the presence or emergence of resistant subpopulations, which may potentially lead to inaccurate antibiotic susceptibility testing results and influence the success of antibiotic therapy.