Influence of patient characteristics on microbial composition in surgical-site infections: insights from national surveillance study.

BACKGROUND: Although the impact of surgery- and patient-dependent factors on surgical-site infections (SSIs) have been studied extensively, their influence on the microbial composition of SSI remains unexplored. The aim of this study was to identify patient-dependent predictors of the microbial composition of SSIs across different types of surgery. METHODS: This retrospective cohort study included 538 893 patients from the Swiss national infection surveillance programme. Multilabel classification methods, adaptive boosting and Gaussian Naive Bayes were employed to identify predictors of the microbial composition of SSIs using 20 features, including sex, age, BMI, duration of surgery, type of surgery, and surgical antimicrobial prophylaxis. RESULTS: Overall, SSIs were recorded in 18 642 patients (3.8%) and, of these, 10 632 had microbiological wound swabs available. The most common pathogens identified in SSIs were Enterobacterales (57%), Staphylococcus spp. (31%), and Enterococcus spp. (28%). Age (mean feature importance 0.260, 95% c.i. 0.209 to 0.309), BMI (0.224, 0.177 to 0.271), and duration of surgery (0.221, 0.180 to 0.269) were strong and independent predictors of the microbial composition of SSIs. Increasing age and duration of surgical procedure as well as decreasing BMI were associated with a shift from Staphylococcus spp. to Enterobacterales and Enterococcus spp. An online application of the machine learning model is available for validation in other healthcare systems. CONCLUSION: Age, BMI, and duration of surgery were key predictors of the microbial composition of SSI, irrespective of the type of surgery, demonstrating the relevance of patient-dependent factors to the pathogenesis of SSIs.

Local infections are a frequent problem after surgery. The risk factors for surgical infections have been identified, but it is unclear which factors predict the type of microorganisms found in such infections. The aim of the present study was to assess patient factors affecting the composition of microorganisms in surgical infections. Data from 538 893 patients were analysed using standard statistics and machine learning methods. The results showed that age, BMI, and the duration of surgery were important in determining the bacteria found in the surgical-site infections. With increasing age, longer operations, and lower BMI, more bacteria stemming from the intestine were found in the surgical site, as opposed to bacteria from the skin. This knowledge may help in developing more personalized treatments for patients undergoing surgery in the future.

Functional identification of cis-regulatory long noncoding RNAs at controlled false discovery rates.

A key attribute of some long noncoding RNAs (lncRNAs) is their ability to regulate expression of neighbouring genes in cis. However, such 'cis-lncRNAs' are presently defined using ad hoc criteria that, we show, are prone to false-positive predictions. The resulting lack of cis-lncRNA catalogues hinders our understanding of their extent, characteristics and mechanisms. Here, we introduce TransCistor, a framework for defining and identifying cis-lncRNAs based on enrichment of targets amongst proximal genes. TransCistor's simple and conservative statistical models are compatible with functionally defined target gene maps generated by existing and future technologies. Using transcriptome-wide perturbation experiments for 268 human and 134 mouse lncRNAs, we provide the first large-scale survey of cis-lncRNAs. Known cis-lncRNAs are correctly identified, including XIST, LINC00240 and UMLILO, and predictions are consistent across analysis methods, perturbation types and independent experiments. We detect cis-activity in a minority of lncRNAs, primarily involving activators over repressors. Cis-lncRNAs are detected by both RNA interference and antisense oligonucleotide perturbations. Mechanistically, cis-lncRNA transcripts are observed to physically associate with their target genes and are weakly enriched with enhancer elements. In summary, TransCistor establishes a quantitative foundation for cis-lncRNAs, opening a path to elucidating their molecular mechanisms and biological significance.

Tumour mutations in long noncoding RNAs enhance cell fitness.

Long noncoding RNAs (lncRNAs) are linked to cancer via pathogenic changes in their expression levels. Yet, it remains unclear whether lncRNAs can also impact tumour cell fitness via function-altering somatic "driver" mutations. To search for such driver-lncRNAs, we here perform a genome-wide analysis of fitness-altering single nucleotide variants (SNVs) across a cohort of 2583 primary and 3527 metastatic tumours. The resulting 54 mutated and positively-selected lncRNAs are significantly enriched for previously-reported cancer genes and a range of clinical and genomic features. A number of these lncRNAs promote tumour cell proliferation when overexpressed in in vitro models. Our results also highlight a dense SNV hotspot in the widely-studied NEAT1 oncogene. To directly evaluate the functional significance of NEAT1 SNVs, we use in cellulo mutagenesis to introduce tumour-like mutations in the gene and observe a significant and reproducible increase in cell fitness, both in vitro and in a mouse model. Mechanistic studies reveal that SNVs remodel the NEAT1 ribonucleoprotein and boost subnuclear paraspeckles. In summary, this work demonstrates the utility of driver analysis for mapping cancer-promoting lncRNAs, and provides experimental evidence that somatic mutations can act through lncRNAs to enhance pathological cancer cell fitness.

Multi-hallmark long noncoding RNA maps reveal non-small cell lung cancer vulnerabilities.

Long noncoding RNAs (lncRNAs) are widely dysregulated in cancer, yet their functional roles in cancer hallmarks remain unclear. We employ pooled CRISPR deletion to perturb 831 lncRNAs detected in KRAS-mutant non-small cell lung cancer (NSCLC) and measure their contribution to proliferation, chemoresistance, and migration across two cell backgrounds. Integrative analysis of these data outperforms conventional "dropout" screens in identifying cancer genes while prioritizing disease-relevant lncRNAs with pleiotropic and background-independent roles. Altogether, 80 high-confidence oncogenic lncRNAs are active in NSCLC, which tend to be amplified and overexpressed in tumors. A follow-up antisense oligonucleotide (ASO) screen shortlisted two candidates, Cancer Hallmarks in Lung LncRNA 1 (CHiLL1) and GCAWKR, whose knockdown consistently suppressed cancer hallmarks in two- and three-dimension tumor models. Molecular phenotyping reveals that CHiLL1 and GCAWKR control cellular-level phenotypes via distinct transcriptional networks. This work reveals a multi-dimensional functional lncRNA landscape underlying NSCLC that contains potential therapeutic vulnerabilities.

Enhancing CRISPR deletion via pharmacological delay of DNA-PKcs.

CRISPR-Cas9 deletion (CRISPR-del) is the leading approach for eliminating DNA from mammalian cells and underpins a variety of genome-editing applications. Target DNA, defined by a pair of double-strand breaks (DSBs), is removed during nonhomologous end-joining (NHEJ). However, the low efficiency of CRISPR-del results in laborious experiments and false-negative results. By using an endogenous reporter system, we show that repression of the DNA-dependent protein kinase catalytic subunit (DNA-PKcs)-an early step in NHEJ-yields substantial increases in DNA deletion. This is observed across diverse cell lines, gene delivery methods, commercial inhibitors, and guide RNAs, including those that otherwise display negligible activity. We further show that DNA-PKcs inhibition can be used to boost the sensitivity of pooled functional screens and detect true-positive hits that would otherwise be overlooked. Thus, delaying the kinetics of NHEJ relative to DSB formation is a simple and effective means of enhancing CRISPR-deletion.

Automatic construction of molecular similarity networks for visual graph mining in chemical space of bioactive peptides: an unsupervised learning approach.

The increasing interest in bioactive peptides with therapeutic potentials has been reflected in a large variety of biological databases published over the last years. However, the knowledge discovery process from these heterogeneous data sources is a nontrivial task, becoming the essence of our research endeavor. Therefore, we devise a unified data model based on molecular similarity networks for representing a chemical reference space of bioactive peptides, having an implicit knowledge that is currently not explicitly accessed in existing biological databases. Indeed, our main contribution is a novel workflow for the automatic construction of such similarity networks, enabling visual graph mining techniques to uncover new insights from the "ocean" of known bioactive peptides. The workflow presented here relies on the following sequential steps: (i) calculation of molecular descriptors by applying statistical and aggregation operators on amino acid property vectors; (ii) a two-stage unsupervised feature selection method to identify an optimized subset of descriptors using the concepts of entropy and mutual information; (iii) generation of sparse networks where nodes represent bioactive peptides, and edges between two nodes denote their pairwise similarity/distance relationships in the defined descriptor space; and (iv) exploratory analysis using visual inspection in combination with clustering and network science techniques. For practical purposes, the proposed workflow has been implemented in our visual analytics software tool ( http://mobiosd-hub.com/starpep/ ), to assist researchers in extracting useful information from an integrated collection of 45120 bioactive peptides, which is one of the largest and most diverse data in its field. Finally, we illustrate the applicability of the proposed workflow for discovering central nodes in molecular similarity networks that may represent a biologically relevant chemical space known to date.

Particle size distribution from extinction and absorption data of metallic nanoparticles.

We describe a method for inverting spectroscopic data of the absorption and extinction properties of colloidal samples of resonant particles. We show that, with some prior knowledge, the genetic algorithm employed is able to estimate the probability density function of particle sizes. Since the data are sensitive to the shape and material of the particles, some information about these properties can also be retrieved. The viability of the method is illustrated by inverting numerically generated data, as well as experimental data obtained with specially prepared samples of metallic nanoparticles in aqueous suspension.

Graph-based data integration from bioactive peptide databases of pharmaceutical interest: toward an organized collection enabling visual network analysis.

MOTIVATION: Bioactive peptides have gained great attention in the academy and pharmaceutical industry since they play an important role in human health. However, the increasing number of bioactive peptide databases is causing the problem of data redundancy and duplicated efforts. Even worse is the fact that the available data is non-standardized and often dirty with data entry errors. Therefore, there is a need for a unified view that enables a more comprehensive analysis of the information on this topic residing at different sites. RESULTS: After collecting web pages from a large variety of bioactive peptide databases, we organized the web content into an integrated graph database (starPepDB) that holds a total of 71 310 nodes and 348 505 relationships. In this graph structure, there are 45 120 nodes representing peptides, and the rest of the nodes are connected to peptides for describing metadata. Additionally, to facilitate a better understanding of the integrated data, a software tool (starPep toolbox) has been developed for supporting visual network analysis in a user-friendly way; providing several functionalities such as peptide retrieval and filtering, network construction and visualization, interactive exploration and exporting data options. AVAILABILITY AND IMPLEMENTATION: Both starPepDB and starPep toolbox are freely available at http://mobiosd-hub.com/starpep/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Classification of riboswitch sequences using k-mer frequencies.

Riboswitches are non-coding RNAs that regulate gene expression by altering the structural conformation of mRNA transcripts. Their regulation mechanism might be exploited for interesting biomedical applications such as drug targets and biosensors. A major challenge consists in accurately identifying metabolite-binding RNA switches which are structurally complex and diverse. In this regard, we investigated the classification of 16 riboswitch families using supervised learning algorithms trained solely with sequence-based features. We generated a reduced feature set and proposed a visual representation to explore its components. We induced Support Vector Machine, Random Forest, Naive Bayes, J48, and HyperPipes classifiers with our proposed feature set and tested their performance over independent data. Our best multi-class classifier achieved F-measure values of 0.996 and 0.966 in the training and test phases, respectively, outperforming those of a previous approach. When compared against BLAST, our best classifiers yielded competitive results. This work shows that the classifiers trained with our sequence-based feature set accurately discriminate riboswitches.