Machine learning for predicting neurodegenerative diseases in the general older population: a cohort study.

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.

Allergic airway inflammation delays glioblastoma progression and reinvigorates systemic and local immunity in mice.

BACKGROUND: Numerous patient-based studies have highlighted the protective role of immunoglobulin E-mediated allergic diseases on glioblastoma (GBM) susceptibility and prognosis. However, the mechanisms behind this observation remain elusive. Our objective was to establish a preclinical model able to recapitulate this phenomenon and investigate the role of immunity underlying such protection. METHODS: An immunocompetent mouse model of allergic airway inflammation (AAI) was initiated before intracranial implantation of mouse GBM cells (GL261). RAG1-KO mice served to assess tumor growth in a model deficient for adaptive immunity. Tumor development was monitored by MRI. Microglia were isolated for functional analyses and RNA-sequencing. Peripheral as well as tumor-associated immune cells were characterized by flow cytometry. The impact of allergy-related microglial genes on patient survival was analyzed by Cox regression using publicly available datasets. RESULTS: We found that allergy establishment in mice delayed tumor engraftment in the brain and reduced tumor growth resulting in increased mouse survival. AAI induced a transcriptional reprogramming of microglia towards a pro-inflammatory-like state, uncovering a microglia gene signature, which correlated with limited local immunosuppression in glioma patients. AAI increased effector memory T-cells in the circulation as well as tumor-infiltrating CD4+ T-cells. The survival benefit conferred by AAI was lost in mice devoid of adaptive immunity. CONCLUSION: Our results demonstrate that AAI limits both tumor take and progression in mice, providing a preclinical model to study the impact of allergy on GBM susceptibility and prognosis, respectively. We identify a potentiation of local and adaptive systemic immunity, suggesting a reciprocal crosstalk that orchestrates allergy-induced immune protection against GBM.

DrDimont: explainable drug response prediction from differential analysis of multi-omics networks.

MOTIVATION: While it has been well established that drugs affect and help patients differently, personalized drug response predictions remain challenging. Solutions based on single omics measurements have been proposed, and networks provide means to incorporate molecular interactions into reasoning. However, how to integrate the wealth of information contained in multiple omics layers still poses a complex problem. RESULTS: We present DrDimont, Drug response prediction from Differential analysis of multi-omics networks. It allows for comparative conclusions between two conditions and translates them into differential drug response predictions. DrDimont focuses on molecular interactions. It establishes condition-specific networks from correlation within an omics layer that are then reduced and combined into heterogeneous, multi-omics molecular networks. A novel semi-local, path-based integration step ensures integrative conclusions. Differential predictions are derived from comparing the condition-specific integrated networks. DrDimont's predictions are explainable, i.e. molecular differences that are the source of high differential drug scores can be retrieved. We predict differential drug response in breast cancer using transcriptomics, proteomics, phosphosite and metabolomics measurements and contrast estrogen receptor positive and receptor negative patients. DrDimont performs better than drug prediction based on differential protein expression or PageRank when evaluating it on ground truth data from cancer cell lines. We find proteomic and phosphosite layers to carry most information for distinguishing drug response. AVAILABILITY AND IMPLEMENTATION: DrDimont is available on CRAN: https://cran.r-project.org/package=DrDimont. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Oncolytic H-1 parvovirus binds to sialic acid on laminins for cell attachment and entry.

H-1 parvovirus (H-1PV) is a promising anticancer therapy. However, in-depth understanding of its life cycle, including the host cell factors needed for infectivity and oncolysis, is lacking. This understanding may guide the rational design of combination strategies, aid development of more effective viruses, and help identify biomarkers of susceptibility to H-1PV treatment. To identify the host cell factors involved, we carry out siRNA library screening using a druggable genome library. We identify one crucial modulator of H-1PV infection: laminin γ1 (LAMC1). Using loss- and gain-of-function studies, competition experiments, and ELISA, we validate LAMC1 and laminin family members as being essential to H-1PV cell attachment and entry. H-1PV binding to laminins is dependent on their sialic acid moieties and is inhibited by heparin. We show that laminins are differentially expressed in various tumour entities, including glioblastoma. We confirm the expression pattern of laminin γ1 in glioblastoma biopsies by immunohistochemistry. We also provide evidence of a direct correlation between LAMC1 expression levels and H-1PV oncolytic activity in 59 cancer cell lines and in 3D organotypic spheroid cultures with different sensitivities to H-1PV infection. These results support the idea that tumours with elevated levels of γ1 containing laminins are more susceptible to H-1PV-based therapies.

Patient-derived organoids and orthotopic xenografts of primary and recurrent gliomas represent relevant patient avatars for precision oncology.

Patient-based cancer models are essential tools for studying tumor biology and for the assessment of drug responses in a translational context. We report the establishment a large cohort of unique organoids and patient-derived orthotopic xenografts (PDOX) of various glioma subtypes, including gliomas with mutations in IDH1, and paired longitudinal PDOX from primary and recurrent tumors of the same patient. We show that glioma PDOXs enable long-term propagation of patient tumors and represent clinically relevant patient avatars that retain histopathological, genetic, epigenetic, and transcriptomic features of parental tumors. We find no evidence of mouse-specific clonal evolution in glioma PDOXs. Our cohort captures individual molecular genotypes for precision medicine including mutations in IDH1, ATRX, TP53, MDM2/4, amplification of EGFR, PDGFRA, MET, CDK4/6, MDM2/4, and deletion of CDKN2A/B, PTCH, and PTEN. Matched longitudinal PDOX recapitulate the limited genetic evolution of gliomas observed in patients following treatment. At the histological level, we observe increased vascularization in the rat host as compared to mice. PDOX-derived standardized glioma organoids are amenable to high-throughput drug screens that can be validated in mice. We show clinically relevant responses to temozolomide (TMZ) and to targeted treatments, such as EGFR and CDK4/6 inhibitors in (epi)genetically defined subgroups, according to MGMT promoter and EGFR/CDK status, respectively. Dianhydrogalactitol (VAL-083), a promising bifunctional alkylating agent in the current clinical trial, displayed high therapeutic efficacy, and was able to overcome TMZ resistance in glioblastoma. Our work underscores the clinical relevance of glioma organoids and PDOX models for translational research and personalized treatment studies and represents a unique publicly available resource for precision oncology.

A Network-Based Embedding Method for Drug-Target Interaction Prediction.

Integration of multi-omics and pharmacological data can help researchers understand the impact of drugs on dynamic biological systems. Network-based approaches to such integration explore the interaction of different cellular components and drugs. However, with ever-increasing amounts of data, processing these high-dimensional biological networks requires powerful tools. We investigate whether network embeddings can address this problem by providing an effective method for dimensionality reduction in drug-related networks. A neural network-based embedding method is employed to encode protein-protein, protein-disease, drug-drug and drug-disease networks for the prediction of novel drug-target interactions. We found that drug-target interaction prediction using embeddings of heterogeneous networks as input features performs comparably to state-of-the-art methods, exhibiting an area under the ROC curve of 84%, outperforming methods such as BLM-NII and NetLapRLS, and coming very close to the best performing network methods such as HNM, CMF and DTINet. These encouraging results suggest that further investigation of this approach is warranted.

Temozolomide-Induced RNA Interactome Uncovers Novel LncRNA Regulatory Loops in Glioblastoma.

Resistance to chemotherapy by temozolomide (TMZ) is a major cause of glioblastoma (GBM) recurrence. So far, attempts to characterize factors that contribute to TMZ sensitivity have largely focused on protein-coding genes, and failed to provide effective therapeutic targets. Long noncoding RNAs (lncRNAs) are essential regulators of epigenetic-driven cell diversification, yet, their contribution to the transcriptional response to drugs is less understood. Here, we performed RNA-seq and small RNA-seq to provide a comprehensive map of transcriptome regulation upon TMZ in patient-derived GBM stem-like cells displaying different drug sensitivity. In a search for regulatory mechanisms, we integrated thousands of molecular associations stored in public databases to generate a background "RNA interactome". Our systems-level analysis uncovered a coordinated program of TMZ response reflected by regulatory circuits that involve transcription factors, mRNAs, miRNAs, and lncRNAs. We discovered 22 lncRNAs involved in regulatory loops and/or with functional relevance in drug response and prognostic value in gliomas. Thus, the investigation of TMZ-induced gene networks highlights novel RNA-based predictors of chemosensitivity in GBM. The computational modeling used to identify regulatory circuits underlying drug response and prioritizing gene candidates for functional validation is applicable to other datasets.

Extracellular ATP and CD39 Activate cAMP-Mediated Mitochondrial Stress Response to Promote Cytarabine Resistance in Acute Myeloid Leukemia.

Relapses driven by chemoresistant leukemic cell populations are the main cause of mortality for patients with acute myeloid leukemia (AML). Here, we show that the ectonucleotidase CD39 (ENTPD1) is upregulated in cytarabine-resistant leukemic cells from both AML cell lines and patient samples in vivo and in vitro. CD39 cell-surface expression and activity is increased in patients with AML upon chemotherapy compared with diagnosis, and enrichment in CD39-expressing blasts is a marker of adverse prognosis in the clinics. High CD39 activity promotes cytarabine resistance by enhancing mitochondrial activity and biogenesis through activation of a cAMP-mediated adaptive mitochondrial stress response. Finally, genetic and pharmacologic inhibition of CD39 ecto-ATPase activity blocks the mitochondrial reprogramming triggered by cytarabine treatment and markedly enhances its cytotoxicity in AML cells in vitro and in vivo. Together, these results reveal CD39 as a new residual disease marker and a promising therapeutic target to improve chemotherapy response in AML. SIGNIFICANCE: Extracellular ATP and CD39-P2RY13-cAMP-OxPHOS axis are key regulators of cytarabine resistance, offering a new promising therapeutic strategy in AML.This article is highlighted in the In This Issue feature, p. 1426.

Machine learning-assisted neurotoxicity prediction in human midbrain organoids.

INTRODUCTION: Brain organoids are highly complex multi-cellular tissue proxies, which have recently risen as novel tools to study neurodegenerative diseases such as Parkinson's disease (PD). However, with increasing complexity of the system, usage of quantitative tools becomes challenging. OBJECTIVES: The primary objective of this study was to develop a neurotoxin-induced PD organoid model and to assess the neurotoxic effect on dopaminergic neurons using microscopy-based phenotyping in a high-content fashion. METHODS: We describe a pipeline for a machine learning-based analytical method, allowing for detailed image-based cell profiling and toxicity prediction in brain organoids treated with the neurotoxic compound 6-hydroxydopamine (6-OHDA). RESULTS: We quantified features such as dopaminergic neuron count and neuronal complexity and built a machine learning classifier with the data to optimize data processing strategies and to discriminate between different treatment conditions. We validated the approach with high content imaging data from PD patient derived midbrain organoids. CONCLUSIONS: The here described model is a valuable tool for advanced in vitro PD modeling and to test putative neurotoxic compounds.

Fisetin protects against cardiac cell death through reduction of ROS production and caspases activity.

Myocardial infarction (MI) is a leading cause of death worldwide. Reperfusion is considered as an optimal therapy following cardiac ischemia. However, the promotion of a rapid elevation of O2 levels in ischemic cells produces high amounts of reactive oxygen species (ROS) leading to myocardial tissue injury. This phenomenon is called ischemia reperfusion injury (IRI). We aimed at identifying new and effective compounds to treat MI and minimize IRI. We previously studied heart regeneration following myocardial injury in zebrafish and described each step of the regeneration process, from the day of injury until complete recovery, in terms of transcriptional responses. Here, we mined the data and performed a deep in silico analysis to identify drugs highly likely to induce cardiac regeneration. Fisetin was identified as the top candidate. We validated its effects in an in vitro model of MI/IRI in mammalian cardiac cells. Fisetin enhances viability of rat cardiomyocytes following hypoxia/starvation - reoxygenation. It inhibits apoptosis, decreases ROS generation and caspase activation and protects from DNA damage. Interestingly, fisetin also activates genes involved in cell proliferation. Fisetin is thus a highly promising candidate drug with clinical potential to protect from ischemic damage following MI and to overcome IRI.

A deep neural network approach to predicting clinical outcomes of neuroblastoma patients.

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.

Analysis of correlation-based biomolecular networks from different omics data by fitting stochastic block models.

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.

Deconvolution of transcriptomes and miRNomes by independent component analysis provides insights into biological processes and clinical outcomes of melanoma patients.

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.

Connecting Histopathology Imaging and Proteomics in Kidney Cancer through Machine Learning.

Proteomics data encode molecular features of diagnostic value and accurately reflect key underlying biological mechanisms in cancers. Histopathology imaging is a well-established clinical approach to cancer diagnosis. The predictive relationship between large-scale proteomics and H&E-stained histopathology images remains largely uncharacterized. Here we investigate such associations through the application of machine learning, including deep neural networks, to proteomics and histology imaging datasets generated by the Clinical Proteomic Tumor Analysis Consortium (CPTAC) from clear cell renal cell carcinoma patients. We report robust correlations between a set of diagnostic proteins and predictions generated by an imaging-based classification model. Proteins significantly correlated with the histology-based predictions are significantly implicated in immune responses, extracellular matrix reorganization, and metabolism. Moreover, we showed that the genes encoding these proteins also reliably recapitulate the biological associations with imaging-derived predictions based on strong gene-protein expression correlations. Our findings offer novel insights into the integrative modeling of histology and omics data through machine learning, as well as the methodological basis for new research opportunities in this and other cancer types.

Inhibition of mitochondrial respiration prevents BRAF-mutant melanoma brain metastasis.

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 ß-sitosterol, a well-tolerated and brain-penetrable phytosterol. Here we show that ß-sitosterol attenuates melanoma cell growth in vitro and also inhibits brain metastasis formation in vivo. Functional analyses indicated that the therapeutic potential of ß-sitosterol was linked to mitochondrial interference. Mechanistically, ß-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 ß-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, ß-sitosterol represents a promising adjuvant to BRAF inhibitor therapy in patients with, or at risk for, melanoma brain metastases.

Long-term treatment with valganciclovir improves lentiviral suicide gene therapy of glioblastoma.

BACKGROUND: Suicide gene therapy for malignant gliomas has shown encouraging results in the latest clinical trials. However, prodrug application was most often restricted to short-term treatment (14 days), especially when replication-defective vectors were used. We previously showed that a substantial fraction of herpes simplex virus thymidine kinase (HSV-TK) transduced tumor cells survive ganciclovir (GCV) treatment in an orthotopic glioblastoma (GBM) xenograft model. Here we analyzed whether these TK+ tumor cells are still sensitive to prodrug treatment and whether prolonged prodrug treatment can enhance treatment efficacy. METHODS: Glioma cells positive for TK and green fluorescent protein (GFP) were sorted from xenograft tumors recurring after suicide gene therapy, and their sensitivity to GCV was tested in vitro. GBM xenografts were treated with HSV-TK/GCV, HSV-TK/valganciclovir (valGCV), or HSV-TK/valGCV + erlotinib. Tumor growth was analyzed by MRI, and survival as well as morphological and molecular changes were assessed. RESULTS: TK-GFP+ tumor cells from recurrent xenograft tumors retained sensitivity to GCV in vitro. Importantly, a prolonged period (3 mo) of prodrug administration with valganciclovir (valGCV) resulted in a significant survival advantage compared with short-term (3 wk) application of GCV. Recurrent tumors from the treatment groups were more invasive and less angiogenic compared with primary tumors and showed significant upregulation of epidermal growth factor receptor (EGFR) expression. However, double treatment with the EGFR inhibitor erlotinib did not increase therapeutic efficacy. CONCLUSION: Long-term treatment with valGCV should be considered as a replacement for short-term treatment with GCV in clinical trials of HSV-TK mediated suicide gene therapy.

Stem cell-associated heterogeneity in Glioblastoma results from intrinsic tumor plasticity shaped by the microenvironment.

The identity and unique capacity of cancer stem cells (CSC) to drive tumor growth and resistance have been challenged in brain tumors. Here we report that cells expressing CSC-associated cell membrane markers in Glioblastoma (GBM) do not represent a clonal entity defined by distinct functional properties and transcriptomic profiles, but rather a plastic state that most cancer cells can adopt. We show that phenotypic heterogeneity arises from non-hierarchical, reversible state transitions, instructed by the microenvironment and is predictable by mathematical modeling. Although functional stem cell properties were similar in vitro, accelerated reconstitution of heterogeneity provides a growth advantage in vivo, suggesting that tumorigenic potential is linked to intrinsic plasticity rather than CSC multipotency. The capacity of any given cancer cell to reconstitute tumor heterogeneity cautions against therapies targeting CSC-associated membrane epitopes. Instead inherent cancer cell plasticity emerges as a novel relevant target for treatment.

Artificial intelligence for precision oncology: beyond patient stratification.

The data-driven identification of disease states and treatment options is a crucial challenge for precision oncology. Artificial intelligence (AI) offers unique opportunities for enhancing such predictive capabilities in the lab and the clinic. AI, including its best-known branch of research, machine learning, has significant potential to enable precision oncology well beyond relatively well-known pattern recognition applications, such as the supervised classification of single-source omics or imaging datasets. This perspective highlights key advances and challenges in that direction. Furthermore, it argues that AI's scope and depth of research need to be expanded to achieve ground-breaking progress in precision oncology.