Cardiovascular risk prediction - a systems medicine approach

E-print/Working paper (2023)

Background Guidelines for the prevention of cardiovascular disease (CVD) have recommended the assessment of the total CVD risk by risk scores. Current risk algorithms are low in sensitivity and ... [more ▼]

Background Guidelines for the prevention of cardiovascular disease (CVD) have recommended the assessment of the total CVD risk by risk scores. Current risk algorithms are low in sensitivity and specificity and they have not incorporated emerging risk markers for CVD. We suggest that CVD risk assessment can be still improved. We have developed a long-term risk prediction model of cardiovascular mortality in patients with stable coronary artery disease (CAD) based on newly available machine learning and on an extended dataset of new biomarkers.Methods 2953 participants of the Ludwigshafen Risk and Cardiovascular Health (LURIC) study were included. 184 laboratory and 21 demographic markers were ranked according to their contribution to risk of cardiovascular (CV) mortality using different data mining approaches. A self-learning bioinformatics workflow, including seven different machine learning algorithms, was developed for CV risk prediction. The study population was stratified into patients with and without significant CAD. Thereby, significant CAD was defined as a lumen narrowing of 50 or more in at least one of the coronary segments or a history of definite myocardial infarction. The machine learning models in both subpopulations were compared with established CV risk assessment tools.Results After a follow-up of 10 years, 603 (20.4%) patients died of cardiovascular causes. 95% patients without CAD deceased within ten years and 247 (13.2 %) patients with CAD within 5 years. Overall and in patients without CAD, NT-proBNP (N-terminal pro B-type natriuretic peptide), TnT (Troponin T), estimated cystatin c based GFR (glomerular filtration rate) and age were the highest ranked predictors, while in patients with CAD, NT-proBNP, GFR, CT-proAVP (C-terminal pro arginine vasopressin) and TNT were highest predictive. In the comparison with the FRS, PROCAM and ESC risk scores, the machine learning workflow produced more accurate and robust CV mortality prediction in patients without CAD. Equivalent CV risk prediction was obtained in the CAD subpopulation in comparison with the Marschner risk score. Overall, the existing algorithms in general tend to assign more patients into the medium risk groups, while the machine learning algorithms tend to have a clearer risk/no risk assignment. The framework is available upon request.Conclusion We have developed a fully automated and self-validating computational framework of machine learning techniques using an extensive database of clinical, routinely and non-routinely measured laboratory data. Our framework predicts long-term CV mortality at least as accurate as existing CVD risk scores. A combination of four highly ranked biomarkers and the random forest approach showed the best predictive results. Moreover, a dynamic computational model has several advantages over static CVD risk prediction tools: it is freeware, transparent, variable, transferable and expandable to any population, types of events and time frames. [less ▲]

Ambra1 modulates the tumor immune microenvironment and response to PD-1 blockade in melanoma.

in Journal for immunotherapy of cancer (2023), 11(3),

BACKGROUND: Loss of Ambra1 (autophagy and beclin 1 regulator 1), a multifunctional scaffold protein, promotes the formation of nevi and contributes to several phases of melanoma development. The ... [more ▼]

BACKGROUND: Loss of Ambra1 (autophagy and beclin 1 regulator 1), a multifunctional scaffold protein, promotes the formation of nevi and contributes to several phases of melanoma development. The suppressive functions of Ambra1 in melanoma are mediated by negative regulation of cell proliferation and invasion; however, evidence suggests that loss of Ambra1 may also affect the melanoma microenvironment. Here, we investigate the possible impact of Ambra1 on antitumor immunity and response to immunotherapy. METHODS: This study was performed using an Ambra1-depleted Braf(V600E) /Pten(-/) (-) genetically engineered mouse (GEM) model of melanoma, as well as GEM-derived allografts of Braf(V600E) /Pten(-/) (-) and Braf(V600E) /Pten(-/) (-)/Cdkn2a(-/) (-) tumors with Ambra1 knockdown. The effects of Ambra1 loss on the tumor immune microenvironment (TIME) were analyzed using NanoString technology, multiplex immunohistochemistry, and flow cytometry. Transcriptome and CIBERSORT digital cytometry analyses of murine melanoma samples and human melanoma patients (The Cancer Genome Atlas) were applied to determine the immune cell populations in null or low-expressing AMBRA1 melanoma. The contribution of Ambra1 on T-cell migration was evaluated using a cytokine array and flow cytometry. Tumor growth kinetics and overall survival analysis in Braf(V600E) /Pten(-/) (-)/Cdkn2a(-/) (-) mice with Ambra1 knockdown were evaluated prior to and after administration of a programmed cell death protein-1 (PD-1) inhibitor. RESULTS: Loss of Ambra1 was associated with altered expression of a wide range of cytokines and chemokines as well as decreased infiltration of tumors by regulatory T cells, a subpopulation of T cells with potent immune-suppressive properties. These changes in TIME composition were associated with the autophagic function of Ambra1. In the Braf(V600E) /Pten(-/) (-)/Cdkn2a(-/) (-) model inherently resistant to immune checkpoint blockade, knockdown of Ambra1 led to accelerated tumor growth and reduced overall survival, but at the same time conferred sensitivity to anti-PD-1 treatment. CONCLUSIONS: This study shows that loss of Ambra1 affects the TIME and the antitumor immune response in melanoma, highlighting new functions of Ambra1 in the regulation of melanoma biology. [less ▲]

scFASTCORMICS: A Contextualization Algorithm to Reconstruct Metabolic Multi-Cell Population Models from Single-Cell RNAseq Data

in Metabolites (2022)

Tumours are composed of various cancer cell populations with different mutation profiles, phenotypes and metabolism that cause them to react to drugs in diverse manners. Increasing the resolution of ... [more ▼]

Tumours are composed of various cancer cell populations with different mutation profiles, phenotypes and metabolism that cause them to react to drugs in diverse manners. Increasing the resolution of metabolic models based on single-cell expression data will provide deeper insight into such metabolic differences and improve the predictive power of the models. scFASTCORMICS is a network contextualization algorithm that builds multi-cell population genome-scale models from single-cell RNAseq data. The models contain a subnetwork for each cell population in a tumour, allowing to capture metabolic variations between these clusters. The subnetworks are connected by a union compartment that permits to simulate metabolite exchanges between cell populations in the microenvironment. scFASTCORMICS uses Pareto optimization to simultaneously maximise the compactness, completeness and specificity of the reconstructed metabolic models. scFASTCORMICS is implemented in MATLAB and requires the installation of the COBRA toolbox, rFASTCORMICS and the IBM CPLEX solver. [less ▲]

Machine learning-based prediction of frailty in elderly people : Data from the Berlin Aging Study II (BASE-II)

Poster (2022, October 09)

Frailty is a geriatric medical condition that is highly associated with age and age-related diseases. The multidimensional consequences of frailty are heavily impacting the quality of life, and will ... [more ▼]

Frailty is a geriatric medical condition that is highly associated with age and age-related diseases. The multidimensional consequences of frailty are heavily impacting the quality of life, and will inevitably increase the burden on healthcare systems in the future. Most importantly, the lack of a universal standard to describe, diagnose, or let alone treat frailty, is further complicating the situation in the long-term. Nowadays, more and more frailty assessment tools are being developed on a regional and institutional basis, which is continuing to drive the heterogeneity in the characterization of frailty further apart. Gaining better insights into the underlying causes and pathophysiology of frailty, and how it is developing in patients is, therefore, required to establish strong and accurately tailored response schemes for frail patients, where currently only symptoms are treated. Thus, in this study, we deployed machine learning-based classification and optimization techniques to predict frailty in the Berlin Aging Study II (BASE-II, N=1512, frail=484) and revealed some of the most informative biomedical information to characterize frailty, including new potential biomarkers. Frailty in BASE-II was measured by the Fried et al. 5-item frailty index, composed of the clinical variables grip strength, weight loss, exhaustion, physical activity, and gait. The level of frailty in BASE-II was adapted for binary classification purposes by merging the pre-frail and frail levels as frail. A configurable in-house pipeline was developed for pre-processing the clinical data, predicting the target disease, and determining the most informative subgroup of clinical measurements with regards to frailty. The best prediction power was yielded with resampling and dimensionality reduction techniques using the F-beta-2 score, and was further increased by adding one item of the Fried et al. frailty index. We suggest that a combination of the easy-to-obtain biomedical information on frailty risk factors together with one Fried et al. phenotype information provided by i.e. smart wearable devices (gait, grip strength, . . . ) could significantly improve the frailty prediction power. [less ▲]

The gut microbial metabolite formate exacerbates colorectal cancer progression

in Nature Metabolism (2022)

The gut microbiome is a key player in the immunomodulatory and protumorigenic microenvironment during colorectal cancer (CRC), as different gut-derived bacteria can induce tumour growth. However, the ... [more ▼]

The gut microbiome is a key player in the immunomodulatory and protumorigenic microenvironment during colorectal cancer (CRC), as different gut-derived bacteria can induce tumour growth. However, the crosstalk between the gut microbiome and the host in relation to tumour cell metabolism remains largely unexplored. Here we show that formate, a metabolite produced by the CRC-associated bacterium Fusobacterium nucleatum, promotes CRC development. We describe molecular signatures linking CRC phenotypes with Fusobacterium abundance. Cocultures of F. nucleatum with patient-derived CRC cells display protumorigenic effects, along with a metabolic shift towards increased formate secretion and cancer glutamine metabolism. We further show that microbiome-derived formate drives CRC tumour invasion by triggering AhR signalling, while increasing cancer stemness. Finally, F. nucleatum or formate treatment in mice leads to increased tumour incidence or size, and Th17 cell expansion, which can favour proinflammatory profiles. Moving beyond observational studies, we identify formate as a gut-derived oncometabolite that is relevant for CRC progression. [less ▲]

Construction and contextualization approaches for protein-protein interaction networks.

in Computational and structural biotechnology journal (2022), 20

Protein-protein interaction network (PPIN) analysis is a widely used method to study the contextual role of proteins of interest, to predict novel disease genes, disease or functional modules, and to ... [more ▼]

Protein-protein interaction network (PPIN) analysis is a widely used method to study the contextual role of proteins of interest, to predict novel disease genes, disease or functional modules, and to identify novel drug targets. PPIN-based analysis uses both generic and context-specific networks. Multiple contextualization methodologies have been described, such as shortest-path algorithms, neighborhood-based methods, and diffusion/propagation algorithms. This review discusses these methods, provides intuitive representations of PPIN contextualization, and also examines how the quality of such context-specific networks could be improved by considering additional sources of evidence. As a heuristic, we observe that tasks such as identifying disease genes, drug targets, and protein complexes should consider local neighborhoods, while uncovering disease mechanisms and discovering disease-pathways would gain from diffusion-based construction. [less ▲]

Drug Target Prediction Using Context-Specific Metabolic Models Reconstructed from rFASTCORMICS.

in Methods in Molecular Biology (2022)

Metabolic modeling is a powerful computational tool to analyze metabolism. It has not only been used to identify metabolic rewiring strategies in cancer but also to predict drug targets and candidate ... [more ▼]

Metabolic modeling is a powerful computational tool to analyze metabolism. It has not only been used to identify metabolic rewiring strategies in cancer but also to predict drug targets and candidate drugs for repurposing. Here, we will elaborate on the reconstruction of context-specific metabolic models of cancer using rFASTCORMICS and the subsequent prediction of drugs for repurposing using our drug prediction workflow. [less ▲]

Project-based learning course on metabolic network modelling in computational systems biology.

in PLoS computational biology (2022), 18(1), 1009711

Project-based learning (PBL) is a dynamic student-centred teaching method that encourages students to solve real-life problems while fostering engagement and critical thinking. Here, we report on a PBL ... [more ▼]

Project-based learning (PBL) is a dynamic student-centred teaching method that encourages students to solve real-life problems while fostering engagement and critical thinking. Here, we report on a PBL course on metabolic network modelling that has been running for several years within the Master in Integrated Systems Biology (MISB) at the University of Luxembourg. This 2-week full-time block course comprises an introduction into the core concepts and methods of constraint-based modelling (CBM), applied to toy models and large-scale networks alongside the preparation of individual student projects in week 1 and, in week 2, the presentation and execution of these projects. We describe in detail the schedule and content of the course, exemplary student projects, and reflect on outcomes and lessons learned. PBL requires the full engagement of students and teachers and gives a rewarding teaching experience. The presented course can serve as a role model and inspiration for other similar courses. [less ▲]

Bruceine D Identified as a Drug Candidate against Breast Cancer by a Novel Drug Selection Pipeline and Cell Viability Assay.

in Pharmaceuticals (Basel, Switzerland) (2022), 15(2),

The multi-target effects of natural products allow us to fight complex diseases like cancer on multiple fronts. Unlike docking techniques, network-based approaches such as genome-scale metabolic modelling ... [more ▼]

The multi-target effects of natural products allow us to fight complex diseases like cancer on multiple fronts. Unlike docking techniques, network-based approaches such as genome-scale metabolic modelling can capture multi-target effects. However, the incompleteness of natural product target information reduces the prediction accuracy of in silico gene knockout strategies. Here, we present a drug selection workflow based on context-specific genome-scale metabolic models, built from the expression data of cancer cells treated with natural products, to predict cell viability. The workflow comprises four steps: first, in silico single-drug and drug combination predictions; second, the assessment of the effects of natural products on cancer metabolism via the computation of a dissimilarity score between the treated and control models; third, the identification of natural products with similar effects to the approved drugs; and fourth, the identification of drugs with the predicted effects in pathways of interest, such as the androgen and estrogen pathway. Out of the initial 101 natural products, nine candidates were tested in a 2D cell viability assay. Bruceine D, emodin, and scutellarein showed a dose-dependent inhibition of MCF-7 and Hs 578T cell proliferation with IC(50) values between 0.7 to 65 μM, depending on the drug and cell line. Bruceine D, extracted from Brucea javanica seeds, showed the highest potency. [less ▲]

Stress-induced inflammation evoked by immunogenic cell death is blunted by the IRE1α kinase inhibitor KIRA6 through HSP60 targeting

in Cell Death and Differentiation (2021)

Mounting evidence indicates that immunogenic therapies engaging the unfolded protein response (UPR) following endoplasmic reticulum (ER) stress favor proficient cancer cell-immune interactions, by ... [more ▼]

Mounting evidence indicates that immunogenic therapies engaging the unfolded protein response (UPR) following endoplasmic reticulum (ER) stress favor proficient cancer cell-immune interactions, by stimulating the release of immunomodulatory/ proinflammatory factors by stressed or dying cancer cells. UPR-driven transcription of proinflammatory cytokines/chemokines exert beneficial or detrimental effects on tumor growth and antitumor immunity, but the cell-autonomous machinery governing the cancer cell inflammatory output in response to immunogenic therapies remains poorly defined. Here, we profiled the transcriptome of cancer cells responding to immunogenic or weakly immunogenic treatments. Bioinformatics-driven pathway analysis indicated that immunogenic treatments instigated a NF-κB/AP-1-inflammatory stress response, which dissociated from both cell death and UPR. This stress-induced inflammation was specifically abolished by the IRE1α-kinase inhibitor KIRA6. Supernatants from immunogenic chemotherapy and KIRA6 co-treated cancer cells were deprived of proinflammatory/chemoattractant factors and failed to mobilize neutrophils and induce dendritic cell maturation. Furthermore, KIRA6 significantly reduced the in vivo vaccination potential of dying cancer cells responding to immunogenic chemotherapy. Mechanistically, we found that the anti-inflammatory effect of KIRA6 was still effective in IRE1α-deficient cells, indicating a hitherto unknown off-target effector of this IRE1α-kinase inhibitor. Generation of a KIRA6-clickable photoaffinity probe, mass spectrometry, and co-immunoprecipitation analysis identified cytosolic HSP60 as a KIRA6 off-target in the IKK-driven NF-κB pathway. In sum, our study unravels that HSP60 is a KIRA6-inhibitable upstream regulator of the NF-κB/AP-1-inflammatory stress responses evoked by immunogenic treatments. It also urges caution when interpreting the anti-inflammatory action of IRE1α chemical inhibitors. [less ▲]