In recent years, we have witnessed both artificial intelligence obtaining remarkable results in clinical decision support systems (CDSSs) and explainable artificial intelligence (XAI) improving the interpretability of these models. In turn, this fosters the adoption by medical personnel and improves trustworthiness of CDSSs. Among others, counterfactual explanations prove to be one such XAI technique particularly suitable for the healthcare domain due to its ease of interpretation, even for less technically proficient staff. However, the generation of high-quality counterfactuals relies on generative models for guidance. Unfortunately, training such models requires a huge amount of data that is beyond the means of ordinary hospitals. In this paper, we therefore propose to use federated learning to allow multiple hospitals to jointly train such generative models while maintaining full data privacy. We demonstrate the superiority of our approach compared to locally generated counterfactuals. Moreover, we prove that generative models for counterfactual generation that are trained using federated learning in a suitable environment perform only marginally worse compared to centrally trained ones while offering the benefit of data privacy preservation. Finally, we integrate our method into a prototypical CDSS for treatment recommendation for sepsis patients, thus providing a proof of concept for real-world application as well as insights and sanity checks from clinical application.

Cellular fatty acids of Helicobacter pylori have taxonomic, physiological, and pathogenic implications. However, little is known about the fatty acid composition under various culture conditions. H. pylori is usually grown on blood-supplemented complex media, and the fatty acids in the blood may affect the fatty acids in the cells. In addition, frequently subcultivated laboratory-adapted strains may have properties different from those of fresh clinical isolates, which are culturable only for a limited number of passages. Therefore, the cellular fatty acid profiles of laboratory-adapted strains (LAS) and freshly isolated strains (FIS) were compared after growth on agar that was fatty acid free and growth on blood agar that contained fatty acids. LAS ATCC 43504, 51932, and 700392 and the FIS IMMi 88, 89, and 92, each with <10 subcultures, were cultured in parallel on a fatty acid-free agar (ISAF) and on 5% sheep blood agar (SBA), which contained oleic acid (18:1 9 c ), hexadecanoic acid (16:0), and octadecanoic acid (18:0). ISAF-grown cultures showed no 18:1 9 c and no appreciable differences between the profiles of FIS and LAS. After culture on SBA, the strains showed 18:1 9 c and increased 16:0 and 18:0 content combined with decreased tetradecanoic acid (14:0) content compared to ISAF-grown cells. The changes in the fatty acid profiles were much more pronounced in FIS than in LAS. LAS are obviously characterized by a lower uptake of the fatty acids from the growth medium than FIS. Furthermore, it could be shown that this LAS behavior is most likely a primary strain attribute that is favored under laboratory conditions. The pronounced uptake of fatty acids by strains with FIS behavior may be associated with the expression of virulence properties.

Recently, a new species, Serratia sarumanii, was described, belonging to a group of strains previously identified as Serratia marcescens in routine clinical analyses. It was shown that the identification of S. marcescens isolates by biochemical testing, mass spectrometry, or 16S rRNA gene sequencing was insufficient to resolve the ‘S. marcescens complex’, while sampling point analysis revealed that many genomes assigned to the S. sarumanii cluster were associated with a clinical context. Thus, here the clinical relevance and local as well as global distribution of S. sarumanii is analyzed. In total, 21 strains from three hospitals in Eastern Westphalia-Lippe (OWL), previously identified as S. marcescens and potential causative agents from severe bacterial infections, were analyzed by genome sequencing and species identification. It could be shown that only one isolate was confirmed as S. marcescens, whereas 10 of the 21 isolates were identified as S. sarumanii, indicating that S. sarumanii is the dominant representative of the “Serratia marcescens” complex in hospitals in OWL. To analyze the global species distribution, all Serratia genomes available in GenBank were reclassified. About one-third of these genomes were identified as S. sarumanii, indicating S. sarumanii as the most dominant Serratia species in clinical settings around the world.

Recently, the new species Serratia sarumanii was described, belonging to a group of strains previously identified as Serratia marcescens in routine clinical analyses. It was shown that the identification of S. marcescens isolates by biochemical testing, mass spectrometry or 16S rRNA gene sequencing was insufficient to resolve the ‘S. marcescens complex’, while sampling point analysis revealed that many genomes assigned to the S. sarumanii cluster were associated with a clinical context. Thus, here the clinical relevance and local as well as global distribution of S. sarumanii is analyzed. In total, 21 strains from three hospitals in Eastern Westphalia-Lippe (OWL), previously identified as S. marcescens and potential causative agents from severe bacterial infections, were analyzed by genome sequencing and species identification. It could be shown that only one isolate was confirmed as S. marcescens, whereas 10 of the 21 isolates were identified as S. sarumanii, indicating that S. sarumanii is the dominant representative of the “Serratia marcescens” complex in hospitals in OWL. To analyze the global species distribution, all Serratia genomes available in GenBank were re-classified. About one third of these genomes were identified as S. sarumanii, indicating S. sarumanii as the most dominant Serratia species in clinical settings around the world.

Since the beginning of the global SARS-CoV-2 pandemic, there have been a number of efforts to understand the mutations and clusters of genetic lines of the SARS-CoV-2 virus. Until now, phylogenetic analysis methods have been used for this purpose. Here we show that Principal Component Analysis (PCA), which is widely used in population genetics, can not only help us to understand existing findings about the mutation processes of the virus, but can also provide even deeper insights into these processes while being less sensitive to sequencing gaps. Here we describe a comprehensive analysis of a 46,046 SARS-CoV-2 genome sequence dataset downloaded from the GISAID database in June of this year. Summary PCA provides deep insights into the analysis of large data sets of SARS-CoV-2 genomes, revealing virus lineages that have thus far been unnoticed.