Search

Abstract Background In the older general population, neurodegenerative diseases (NDs) are associated with increased disability, decreased physical and cognitive function. Detecting risk factors can help implement prevention measures. Using deep neural networks (DNNs), a machine-learning algorithm could be an alternative to Cox regression in tabular datasets with many predictive features. We aimed to compare the performance of different types of DNNs with regularized Cox proportional hazards models to predict NDs in the older general population. Methods We performed a longitudinal analysis with participants of the English Longitudinal Study of Ageing. We included men and women with no NDs at baseline, aged 60 years and older, assessed every 2 years from 2004 to 2005 (wave2) to 2016–2017 (wave 8). The features were a set of 91 epidemiological and clinical baseline variables. The outcome was new events of Parkinson’s, Alzheimer or dementia. After applying multiple imputations, we trained three DNN algorithms: Feedforward, TabTransformer, and Dense Convolutional (Densenet). In addition, we trained two algorithms based on Cox models: Elastic Net regularization (CoxEn) and selected features (CoxSf). Results 5433 participants were included in wave 2. During follow-up, 12.7% participants developed NDs. Although the five models predicted NDs events, the discriminative ability was superior using TabTransformer (Uno’s C-statistic (coefficient (95% confidence intervals)) 0.757 (0.702, 0.805). TabTransformer showed superior time-dependent balanced accuracy (0.834 (0.779, 0.889)) and specificity (0.855 (0.0.773, 0.909)) than the other models. With the CoxSf (hazard ratio (95% confidence intervals)), age (10.0 (6.9, 14.7)), poor hearing (1.3 (1.1, 1.5)) and weight loss 1.3 (1.1, 1.6)) were associated with a higher DNN risk. In contrast, executive function (0.3 (0.2, 0.6)), memory (0, 0, 0.1)), increased gait speed (0.2, (0.1, 0.4)), vigorous physical activity (0.7, 0.6, 0.9)) and higher BMI (0.4 (0.2, 0.8)) were associated with a lower DNN risk. Conclusion TabTransformer is promising for prediction of NDs with heterogeneous tabular datasets with numerous features. Moreover, it can handle censored data. However, Cox models perform well and are easier to interpret than DNNs. Therefore, they are still a good choice for NDs.

Rat H-1 parvovirus (H-1PV) is in clinical development for oncolytic therapy. Here, Kulkarni et al. identify LAMC1 as a modulator of H-1PV cell attachment and entry and find that LAMC1 levels and H-1PV oncolytic activity correlate in 59 tested cancer cell lines.

Abstract Background The availability of high-throughput omics datasets from large patient cohorts has allowed the development of methods that aim at predicting patient clinical outcomes, such as survival and disease recurrence. Such methods are also important to better understand the biological mechanisms underlying disease etiology and development, as well as treatment responses. Recently, different predictive models, relying on distinct algorithms (including Support Vector Machines and Random Forests) have been investigated. In this context, deep learning strategies are of special interest due to their demonstrated superior performance over a wide range of problems and datasets. One of the main challenges of such strategies is the “small n large p” problem. Indeed, omics datasets typically consist of small numbers of samples and large numbers of features relative to typical deep learning datasets. Neural networks usually tackle this problem through feature selection or by including additional constraints during the learning process. Methods We propose to tackle this problem with a novel strategy that relies on a graph-based method for feature extraction, coupled with a deep neural network for clinical outcome prediction. The omics data are first represented as graphs whose nodes represent patients, and edges represent correlations between the patients’ omics profiles. Topological features, such as centralities, are then extracted from these graphs for every node. Lastly, these features are used as input to train and test various classifiers. Results We apply this strategy to four neuroblastoma datasets and observe that models based on neural networks are more accurate than state of the art models (DNN: 85%-87%, SVM/RF: 75%-82%). We explore how different parameters and configurations are selected in order to overcome the effects of the small data problem as well as the curse of dimensionality. Conclusions Our results indicate that the deep neural networks capture complex features in the data that help predicting patient clinical outcomes.

Abstract Background The amount of publicly available cancer-related “omics” data is constantly growing and can potentially be used to gain insights into the tumour biology of new cancer patients, their diagnosis and suitable treatment options. However, the integration of different datasets is not straightforward and requires specialized approaches to deal with heterogeneity at technical and biological levels. Methods Here we present a method that can overcome technical biases, predict clinically relevant outcomes and identify tumour-related biological processes in patients using previously collected large discovery datasets. The approach is based on independent component analysis (ICA) – an unsupervised method of signal deconvolution. We developed parallel consensus ICA that robustly decomposes transcriptomics datasets into expression profiles with minimal mutual dependency. Results By applying the method to a small cohort of primary melanoma and control samples combined with a large discovery melanoma dataset, we demonstrate that our method distinguishes cell-type specific signals from technical biases and allows to predict clinically relevant patient characteristics. We showed the potential of the method to predict cancer subtypes and estimate the activity of key tumour-related processes such as immune response, angiogenesis and cell proliferation. ICA-based risk score was proposed and its connection to patient survival was validated with an independent cohort of patients. Additionally, through integration of components identified for mRNA and miRNA data, the proposed method helped deducing biological functions of miRNAs, which would otherwise not be possible. Conclusions We present a method that can be used to map new transcriptomic data from cancer patient samples onto large discovery datasets. The method corrects technical biases, helps characterizing activity of biological processes or cell types in the new samples and provides the prognosis of patient survival.

Cancer stem cells (CSCs) comprise a putative population that can drive growth and resistance. Here, in glioblastoma models the authors show that rather than being a distinct clonal entity, the CSC population represents a plastic state adoptable by most cancer cells via reversible state transitions induced by the microenvironment.

Abstract Melanoma patients carry a high risk of developing brain metastases, and improvements in survival are still measured in weeks or months. Durable disease control within the brain is impeded by poor drug penetration across the blood-brain barrier, as well as intrinsic and acquired drug resistance. Augmented mitochondrial respiration is a key resistance mechanism in BRAF-mutant melanomas but, as we show in this study, this dependence on mitochondrial respiration may also be exploited therapeutically. We first used high-throughput pharmacogenomic profiling to identify potentially repurposable compounds against BRAF-mutant melanoma brain metastases. One of the compounds identified was β-sitosterol, a well-tolerated and brain-penetrable phytosterol. Here we show that β-sitosterol attenuates melanoma cell growth in vitro and also inhibits brain metastasis formation in vivo. Functional analyses indicated that the therapeutic potential of β-sitosterol was linked to mitochondrial interference. Mechanistically, β-sitosterol effectively reduced mitochondrial respiratory capacity, mediated by an inhibition of mitochondrial complex I. The net result of this action was increased oxidative stress that led to apoptosis. This effect was only seen in tumor cells, and not in normal cells. Large-scale analyses of human melanoma brain metastases indicated a significant role of mitochondrial complex I compared to brain metastases from other cancers. Finally, we observed completely abrogated BRAF inhibitor resistance when vemurafenib was combined with either β-sitosterol or a functional knockdown of mitochondrial complex I. In conclusion, based on its favorable tolerability, excellent brain bioavailability, and capacity to inhibit mitochondrial respiration, β-sitosterol represents a promising adjuvant to BRAF inhibitor therapy in patients with, or at risk for, melanoma brain metastases.

Abstract Background One of the main current challenges in computational biology is to make sense of the huge amounts of multidimensional experimental data that are being produced. For instance, large cohorts of patients are often screened using different high-throughput technologies, effectively producing multiple patient-specific molecular profiles for hundreds or thousands of patients. Results We propose and implement a network-based method that integrates such patient omics data into Patient Similarity Networks. Topological features derived from these networks were then used to predict relevant clinical features. As part of the 2017 CAMDA challenge, we have successfully applied this strategy to a neuroblastoma dataset, consisting of genomic and transcriptomic data. In particular, we observe that models built on our network-based approach perform at least as well as state of the art models. We furthermore explore the effectiveness of various topological features and observe, for instance, that redundant centrality metrics can be combined to build more powerful models. Conclusion We demonstrate that the networks inferred from omics data contain clinically relevant information and that patient clinical outcomes can be predicted using only network topological data. Reviewers This article was reviewed by Yang-Yu Liu, Tomislav Smuc and Isabel Nepomuceno.

Background: Biological entities such as genes, promoters, mRNA, metabolites or proteins do not act alone, but in concert in their network context. Modules, i.e., groups of nodes with similar topological properties in these networks characterize important biological functions of the underlying biomolecular system. Edges in such molecular networks represent regulatory and physical interactions, and comparing them between conditions provides valuable information on differential molecular mechanisms. However, biological data is inherently noisy and network reduction techniques can propagate errors particularly to the level of edges. We aim to improve the analysis of networks of biological molecules by deriving modules together with edge relevance estimations that are based on global network characteristics. Methods: The key challenge we address here is investigating the capability of stochastic block models (SBMs) for representing and analyzing different types of biomolecular networks. Fitting them to SBMs both delivers modules of the networks and enables the derivation of edge confidence scores, and it has not yet been investigated for analyzing biomolecular networks. We apply SBM-based analysis independently to three correlation-based networks of breast cancer data originating from high-throughput measurements of different molecular layers: either transcriptomics, proteomics, or metabolomics. The networks were reduced by thresholding for correlation significance or by requirements on scale-freeness. Results and discussion: We find that the networks are best represented by the hierarchical version of the SBM, and many of the predicted blocks have a biologically and phenotypically relevant functional annotation. The edge confidence scores are overall in concordance with the biological evidence given by the measurements. We conclude that biomolecular networks can be appropriately represented and analyzed by fitting SBMs. As the SBM-derived edge confidence scores are based on global network connectivity characteristics and potential hierarchies within the biomolecular networks are considered, they could be used as additional, integrated features in network-based data comparisons.

Background: The topological analysis of networks extracted from different types of “omics” data is a useful strategy for characterizing biologically meaningful properties of the complex systems underlying these networks. In particular, the biological significance of highly connected genes in diverse molecular networks has been previously determined using data from several model organisms and phenotypes. Despite such insights, the predictive potential of candidate hubs in gene co-expression networks in the specific context of cancer-related drug experiments remains to be deeply investigated. The examination of such associations may offer opportunities for the accurate prediction of anticancer drug responses. Methods: Here, we address this problem by: a) analyzing a co-expression network obtained from thousands of cancer cell lines, b) detecting significant network hubs, and c) assessing their capacity to predict drug sensitivity using data from thousands of drug experiments. We investigated the prediction capability of those genes using a multiple linear regression model, independent datasets, comparisons with other models and our own in vitro experiments. Results: These analyses led to the identification of 47 hub genes, which are implicated in a diverse range of cancer-relevant processes and pathways. Overall, encouraging agreements between predicted and observed drug sensitivities were observed in public datasets, as well as in our in vitro validations for four glioblastoma cell lines and four drugs. To facilitate further research, we share our hub-based drug sensitivity prediction model as an online tool. Conclusions: Our research shows that co-expression network hubs are biologically interesting and exhibit potential for predicting drug responses in vitro. These findings motivate further investigations about the relevance and application of our unbiased discovery approach in pre-clinical, translationally-oriented research.

Abstract Long noncoding RNAs (lncRNAs) are emerging as regulators of gene expression in pathogenesis, including cancer. Recently, lncRNAs have been implicated in progression of specific subtypes of breast cancer. One aggressive, basal‐like subtype associates with increased EGFR signaling, while another, the HER2‐enriched subtype, engages a kin of EGFR. Based on the premise that EGFR‐regulated lncRNAs might control the aggressiveness of basal‐like tumors, we identified multiple EGFR‐inducible lncRNAs in basal‐like normal cells and overlaid them with the transcriptomes of over 3,000 breast cancer patients. This led to the identification of 11 prognostic lncRNAs. Functional analyses of this group uncovered LINC01089 (here renamed LncRNA Inhibiting Metastasis; LIMT), a highly conserved lncRNA, which is depleted in basal‐like and in HER2‐positive tumors, and the low expression of which predicts poor patient prognosis. Interestingly, EGF rapidly downregulates LIMT expression by enhancing histone deacetylation at the respective promoter. We also find that LIMT inhibits extracellular matrix invasion of mammary cells in vitro and tumor metastasis in vivo. In conclusion, lncRNAs dynamically regulated by growth factors might act as novel drivers of cancer progression and serve as prognostic biomarkers.

Abstract The zebrafish has the capacity to regenerate its heart after severe injury. While the function of a few genes during this process has been studied, we are far from fully understanding how genes interact to coordinate heart regeneration. To enable systematic insights into this phenomenon, we generated and integrated a dynamic co-expression network of heart regeneration in the zebrafish and linked systems-level properties to the underlying molecular events. Across multiple post-injury time points, the network displays topological attributes of biological relevance. We show that regeneration steps are mediated by modules of transcriptionally coordinated genes, and by genes acting as network hubs. We also established direct associations between hubs and validated drivers of heart regeneration with murine and human orthologs. The resulting models and interactive analysis tools are available at http://infused.vital-it.ch . Using a worked example, we demonstrate the usefulness of this unique open resource for hypothesis generation and in silico screening for genes involved in heart regeneration.

We report that CD47 was upregulated in different EMT-activated human breast cancer cells versus epithelial MCF7 cells. Overexpression of SNAI1 or ZEB1 in epithelial MCF7 cells activated EMT and upregulated CD47 while siRNA-mediated targeting of SNAI1 or ZEB1 in mesenchymal MDA-MB-231 cells reversed EMT and strongly decreased CD47. Mechanistically, SNAI1 and ZEB1 upregulated CD47 by binding directly to E-boxes in the human CD47 promoter. TCGA and METABRIC data sets from breast cancer patients revealed that CD47 correlated with SNAI1 and Vimentin. At functional level, different EMT-activated breast cancer cells were less efficiently phagocytosed by macrophages vs. MCF7 cells. The phagocytosis of EMT-activated cells was rescued by using CD47 blocking antibody or by genetic targeting of SNAI1, ZEB1 or CD47. These results provide a rationale for an innovative preclinical combination immunotherapy based on PD-1/PD-L1 and CD47 blockade along with EMT inhibitors in patients with highly aggressive, mesenchymal, and metastatic breast cancer.

Matrix metalloproteinase 9 (MMP9) is functionally implicated in the process of infarct healing. Several genetic variation of the MMP9 gene have been described, among which the MMP9 Arg668Gln polymorphism. In the present study, we assessed whether this polymorphism influences outcome after acute myocardial infarction (MI). One thousand forty-nine patients undergoing coronary angiography were genotyped for the MMP9 Arg668Gln polymorphism by TaqMan allelic discrimination assay. This population included 154 controls, 161 patients with non ST-elevation MI (NSTEMI), 504 patients with ST-elevation MI (STEMI), and 230 patients with angina. Frequency of the MMP9 Arg668Gln polymorphism in the global population was 25.1%, and was comparable between all groups. STEMI patients had higher creatine phosphokinase (CPK), troponin T (TnT) and MMP9 plasma levels and had lower ejection fraction (EF) than NSTEMI patients. However, the polymorphism was not associated with infarct severity as determined by peak CPK and TnT levels, nor with LV remodeling and outcome as assessed by 1-month EF and NYHA class, as well as 2- year mortality. In silico molecular modeling simulations predicted that the MMP9 polymorphism may decrease MMP9 activity, but this could not be verified by plasma determinations. This study investigated for the first time the association between the MMP9 Arg668Gln polymorphism and clinical outcome after acute MI. Our results indicate that the polymorphism does not seem to be associated with clinical outcome and in particular with the development of left ventricular dysfunction and heart failure.

Objective. Determine the potential of procalcitonin (PCT) to predict neurological outcome after hypothermia treatment following cardiac arrest. Methods. Retrospective analysis of patient data over a 2-year period. Mortality and neurological outcome of survivors were determined 6 months after cardiac arrest using the Cerebral Performance Category (CPC) score. Results. Data from 53 consecutive patients were analyzed. Median age was 63 (54–71) and 79% were male. Twenty-seven patients had good outcome (CPC ≤ 2) whereas 26 had severe neurological sequelae or died (CPC 3–5). At 48 h, after regaining normothermia, PCT was significantly higher in patients with bad outcome compared to those with good outcome: 3.38 (1.10–24.48) versus 0.28 (0–0.75) ng/mL (𝑃

BACKGROUND: A significant proportion of myocardial infarction (MI) patients undergo complex, coordinated perturbations at the molecular level that may eventually drive the occurrence of ventricular dysfunction and heart failure. Despite advances in the elucidation of key processes implicated in this condition, traditional methods relying on gene expression data and the identification of individual biomarkers in isolation pose major limitations not only for improving prediction power, but also for model interpretability. Mechanisms underlying clinical responses after MI remain elusive and there is no biomarker with the capacity to accurately predict ventricular dysfunction after MI. This calls for the exploration of system-level modeling of ventricular dysfunction in post-MI patients. Within this discovery framework key perturbations and predictive patterns are characterized by the integrated biological activity levels observed in pathways, rather than in individual genes. METHODOLOGY/PRINCIPAL FINDINGS: Here we report an integrative approach to identifying pathways related with ventricular dysfunction post MI with potential prognostic and therapeutic value. We found that a diversity of pathway-level perturbations can be profiled in samples of patients with ventricular dysfunction post MI, most of which represent major reductions of gene expression. Highly perturbed pathways included those implicated in antigen-dependent B-cell activation and the synthesis of leucine. By analyzing patient-specific samples encoded with information derived from highly-perturbed pathways, it is possible to visualize differential prognostic patterns and to perform computational classification of patients with areas under the receiver operating characteristic curve above 0.75. We also demonstrate how the integration of the outcomes generated by different pathway-based analysis models may improve ventricular dysfunction prediction performance. SIGNIFICANCE: This research offers an alternative, comprehensive view of key relationships and perturbations that may trigger the emergence or prevention of ventricular dysfunction post-MI.

BACKGROUND: Gene expression profiling and the analysis of protein-protein interaction (PPI) networks may support the identification of disease bio-markers and potential drug targets. Thus, a step forward in the development of systems approaches to medicine is the integrative analysis of these data sources in specific pathological conditions. We report such an integrative bioinformatics analysis in human heart failure (HF). A global PPI network in HF was assembled, which by itself represents a useful compendium of the current status of human HF-relevant interactions. This provided the basis for the analysis of interaction connectivity patterns in relation to a HF gene expression data set. RESULTS: Relationships between the significance of the differentiation of gene expression and connectivity degrees in the PPI network were established. In addition, relationships between gene co-expression and PPI network connectivity were analysed. Highly-connected proteins are not necessarily encoded by genes significantly differentially expressed. Genes that are not significantly differentially expressed may encode proteins that exhibit diverse network connectivity patterns. Furthermore, genes that were not defined as significantly differentially expressed may encode proteins with many interacting partners. Genes encoding network hubs may exhibit weak co-expression with the genes encoding their interacting protein partners. We also found that hubs and superhubs display a significant diversity of co-expression patterns in comparison to peripheral nodes. Gene Ontology (GO) analysis established that highly-connected proteins are likely to be engaged in higher level GO biological process terms, while low-connectivity proteins tend to be engaged in more specific disease-related processes. CONCLUSION: This investigation supports the hypothesis that the integrative analysis of differential gene expression and PPI network analysis may facilitate a better understanding of functional roles and the identification of potential drug targets in human heart failure.

Background: Biological entities such as genes, promoters, mRNA, metabolites or proteins do not act alone, but in concert in their network context. Modules, i.e., groups of nodes with similar topological properties in these networks characterize important biological functions of the underlying biomolecular system. Edges in such molecular networks represent regulatory and physical interactions, and comparing them between conditions provides valuable information on differential molecular mechanisms. However, biological data is inherently noisy and network reduction techniques can propagate errors particularly to the level of edges. We aim to improve the analysis of networks of biological molecules by deriving modules together with edge relevance estimations that are based on global network characteristics. Methods: We propose to fit the networks to stochastic block models (SBM), a method that has not yet been investigated for the analysis of biomolecular networks. This procedure both delivers modules of the networks and enables the derivation of edge confidence scores. We apply it to correlation-based networks of breast cancer data originating from high-throughput measurements of diverse molecular layers such as transcriptomics, proteomics, and metabolomics. The networks were reduced by thresholding for correlation significance or by requirements on scale-freeness. Results and discussion: We find that the networks are best represented by the hierarchical version of the SBM, and many of the predicted blocks have a biological meaning according to functional annotation. The edge confidence scores are overall in concordance with the biological evidence given by the measurements. As they are based on global network connectivity characteristics and potential hierarchies within the biomolecular networks are taken into account, they could be used as additional, integrated features in network-based data comparisons. Their tight relationship to edge existence probabilities can be exploited to predict missing or spurious edges in order to improve the network representation of the underlying biological system.

Background: The topological analysis of networks extracted from different types of “omics” data is a useful strategy for characterizing biologically meaningful properties of the complex systems underlying these networks. In particular, the biological significance of highly connected genes in diverse molecular networks has been previously determined using data from several model organisms and phenotypes. Despite such insights, the predictive potential of candidate hubs in gene co-expression networks in the specific context of cancer-related drug experiments remains to be deeply investigated. The examination of such associations may offer opportunities for the accurate prediction of anticancer drug responses. Methods: Here, we address this problem by: a) analyzing a co-expression network obtained from thousands of cancer cell lines, b) detecting significant network hubs, and c) assessing their capacity to predict drug sensitivity using data from thousands of drug experiments. We investigated the prediction capability of those genes using a multiple linear regression model, independent datasets, comparisons with other models and our own in vitro experiments. Results: These analyses led to the identification of 47 hub genes, which are implicated in a diverse range of cancer-relevant processes and pathways. Overall, encouraging agreements between predicted and observed drug sensitivities were observed in public datasets, as well as in our in vitro validations for four glioblastoma cell lines and four drugs. To facilitate further research, we share our hub-based drug sensitivity prediction model as an online tool. Conclusions: Our research shows that co-expression network hubs are biologically interesting and exhibit potential for predicting drug responses in vitro. These findings motivate further investigations about the relevance and application of our unbiased discovery approach in pre-clinical, translationally-oriented research.