Search

Autoencoders are frequently used to embed molecules for training of downstream deep learning models. However, evaluation of the chemical information quality in the latent spaces is lacking and the model architectures are often arbitrarily chosen. Unoptimized architectures may not only negatively affect latent space quality but also increase energy consumption during training, making the models unsustainable. We conducted systematic experiments to better understand how the autoencoder architecture affects the reconstruction and latent space quality and how it can be optimized towards the encoding task as well as energy consumption. We can show that optimizing the architecture allows us to maintain the quality of a generic architecture but using 97% less data and reducing energy consumption by around 36%. We additionally observed that representing the molecules as SELFIES reduced the reconstruction performance compared to SMILES and that training with enumerated SMILES drastically improved latent space quality. Scientific Contribution: This work provides the first comprehensive systematic analysis of how choosing the autoencoder architecture affects the reconstruction performance of small molecules, the chemical information content of the latent space as well as the energy required for training. Demonstrated on the MOSES benchmarking dataset it provides first valuable insights into how autoencoders for the embedding of small molecules can be designed to optimize their utility and simultaneously become more sustainable, both in terms of energy consumption as well as the required amount of training data. All code, data and model checkpoints are made available on Zenodo (Oestreich et al. Small molecule autoencoders: architecture engineering to optimize latent space utility and sustainability. Zenodo, 2024). Furthermore, the top models can be found on GitHub with scripts to encode custom molecules: https://github.com/MarieOestreich/small-molecule-autoencoders ., (© 2024. The Author(s).)

Background: A total of three intensive care units (ICU) at a German university hospital were involved in an outbreak of Burkholderia cepacia complex (Bcc)., Methods: Patients with microbiological detection of Bcc were evaluated. Products used for mouth hygiene were microbiologically tested. The clonal identity of Bcc was proven by pulse-field gel electrophoresis (PFGE)., Results: On 3 ICUs 12 cases were identified whereby the first detection of Bcc was in respiratory specimens of 11 patients and 1 in a wound swab from the oral cavity. Of these patients six developed ventilator-associated pneumonia (VAP). Investigations revealed that five different batches of an alcohol-free mouthwash containing hexetidine were highly contaminated. Isolates of Bcc from patients and mouthwashes were genetically indistinguishable. A recall of the product was initiated. After elimination of the product from the ICUs no more cases were identified., Conclusions: The source of the outbreak was an intrinsically contaminated alcohol-free mouthwash. Detection of Bcc in specimens from ICU patients should lead to further investigations. Antiseptic oral care products are recommended for reducing the risk of VAP but they may be microbiologically contaminated and, in consequence, increase the risk. The safety of patient care products should be increased by stricter regulations.

Sequencing of bacterial genomes has been progressing with breathtaking speed. Currently, the genomes of 23 bacterial species are sequenced, with approximately 40 more sequencing projects in progress. Industrial research is now facing the challenge of translating this information efficiently into drug discovery. This review will summarize the impact of bacterial genomics, bioinformatics and second-generation genomic technologies on target identification, assay development, lead optimization and compound characterization.