Privacy-Preserving Federated Survival Support Vector Machines for Cross-Institutional Time-To-Event Analysis: Algorithm Development and Validation.

BACKGROUND: Central collection of distributed medical patient data is problematic due to strict privacy regulations. Especially in clinical environments, such as clinical time-to-event studies, large sample sizes are critical but usually not available at a single institution. It has been shown recently that federated learning, combined with privacy-enhancing technologies, is an excellent and privacy-preserving alternative to data sharing. OBJECTIVE: This study aims to develop and validate a privacy-preserving, federated survival support vector machine (SVM) and make it accessible for researchers to perform cross-institutional time-to-event analyses. METHODS: We extended the survival SVM algorithm to be applicable in federated environments. We further implemented it as a FeatureCloud app, enabling it to run in the federated infrastructure provided by the FeatureCloud platform. Finally, we evaluated our algorithm on 3 benchmark data sets, a large sample size synthetic data set, and a real-world microbiome data set and compared the results to the corresponding central method. RESULTS: Our federated survival SVM produces highly similar results to the centralized model on all data sets. The maximal difference between the model weights of the central model and the federated model was only 0.001, and the mean difference over all data sets was 0.0002. We further show that by including more data in the analysis through federated learning, predictions are more accurate even in the presence of site-dependent batch effects. CONCLUSIONS: The federated survival SVM extends the palette of federated time-to-event analysis methods by a robust machine learning approach. To our knowledge, the implemented FeatureCloud app is the first publicly available implementation of a federated survival SVM, is freely accessible for all kinds of researchers, and can be directly used within the FeatureCloud platform.

An overview of food lipids toward food lipidomics.

Increasing evidence regarding lipids' beneficial effects on human health has changed the common perception of consumers and dietary officials about the role(s) of food lipids in a healthy diet. However, lipids are a wide group of molecules with specific nutritional and bioactive properties. To understand their true nutritional and functional value, robust methods are needed for accurate identification and quantification. Specific analytical strategies are crucial to target specific classes, especially the ones present in trace amounts. Finding a unique and comprehensive methodology to cover the full lipidome of each foodstuff is still a challenge. This review presents an overview of the lipids nutritionally relevant in foods and new trends in food lipid analysis for each type/class of lipids. Food lipid classes are described following the LipidMaps classification, fatty acids, endocannabinoids, waxes, C8 compounds, glycerophospholipids, glycerolipids (i.e., glycolipids, betaine lipids, and triglycerides), sphingolipids, sterols, sercosterols (vitamin D), isoprenoids (i.e., carotenoids and retinoids (vitamin A)), quinones (i.e., coenzyme Q, vitamin K, and vitamin E), terpenes, oxidized lipids, and oxylipin are highlighted. The uniqueness of each food group: oil-, protein-, and starch-rich, as well as marine foods, fruits, and vegetables (water-rich) regarding its lipid composition, is included. The effect of cooking, food processing, and storage, in addition to the importance of lipidomics in food quality and authenticity, are also discussed. A critical review of challenges and future trends of the analytical approaches and computational methods in global food lipidomics as the basis to increase consumer awareness of the significant role of lipids in food quality and food security worldwide is presented.

The FeatureCloud Platform for Federated Learning in Biomedicine: Unified Approach.

BACKGROUND: Machine learning and artificial intelligence have shown promising results in many areas and are driven by the increasing amount of available data. However, these data are often distributed across different institutions and cannot be easily shared owing to strict privacy regulations. Federated learning (FL) allows the training of distributed machine learning models without sharing sensitive data. In addition, the implementation is time-consuming and requires advanced programming skills and complex technical infrastructures. OBJECTIVE: Various tools and frameworks have been developed to simplify the development of FL algorithms and provide the necessary technical infrastructure. Although there are many high-quality frameworks, most focus only on a single application case or method. To our knowledge, there are no generic frameworks, meaning that the existing solutions are restricted to a particular type of algorithm or application field. Furthermore, most of these frameworks provide an application programming interface that needs programming knowledge. There is no collection of ready-to-use FL algorithms that are extendable and allow users (eg, researchers) without programming knowledge to apply FL. A central FL platform for both FL algorithm developers and users does not exist. This study aimed to address this gap and make FL available to everyone by developing FeatureCloud, an all-in-one platform for FL in biomedicine and beyond. METHODS: The FeatureCloud platform consists of 3 main components: a global frontend, a global backend, and a local controller. Our platform uses a Docker to separate the local acting components of the platform from the sensitive data systems. We evaluated our platform using 4 different algorithms on 5 data sets for both accuracy and runtime. RESULTS: FeatureCloud removes the complexity of distributed systems for developers and end users by providing a comprehensive platform for executing multi-institutional FL analyses and implementing FL algorithms. Through its integrated artificial intelligence store, federated algorithms can easily be published and reused by the community. To secure sensitive raw data, FeatureCloud supports privacy-enhancing technologies to secure the shared local models and assures high standards in data privacy to comply with the strict General Data Protection Regulation. Our evaluation shows that applications developed in FeatureCloud can produce highly similar results compared with centralized approaches and scale well for an increasing number of participating sites. CONCLUSIONS: FeatureCloud provides a ready-to-use platform that integrates the development and execution of FL algorithms while reducing the complexity to a minimum and removing the hurdles of federated infrastructure. Thus, we believe that it has the potential to greatly increase the accessibility of privacy-preserving and distributed data analyses in biomedicine and beyond.

Lipid network and moiety analysis for revealing enzymatic dysregulation and mechanistic alterations from lipidomics data.

Lipidomics is of growing importance for clinical and biomedical research due to many associations between lipid metabolism and diseases. The discovery of these associations is facilitated by improved lipid identification and quantification. Sophisticated computational methods are advantageous for interpreting such large-scale data for understanding metabolic processes and their underlying (patho)mechanisms. To generate hypothesis about these mechanisms, the combination of metabolic networks and graph algorithms is a powerful option to pinpoint molecular disease drivers and their interactions. Here we present lipid network explorer (LINEX$^2$), a lipid network analysis framework that fuels biological interpretation of alterations in lipid compositions. By integrating lipid-metabolic reactions from public databases, we generate dataset-specific lipid interaction networks. To aid interpretation of these networks, we present an enrichment graph algorithm that infers changes in enzymatic activity in the context of their multispecificity from lipidomics data. Our inference method successfully recovered the MBOAT7 enzyme from knock-out data. Furthermore, we mechanistically interpret lipidomic alterations of adipocytes in obesity by leveraging network enrichment and lipid moieties. We address the general lack of lipidomics data mining options to elucidate potential disease mechanisms and make lipidomics more clinically relevant.

Federated machine learning for a facilitated implementation of Artificial Intelligence in healthcare - a proof of concept study for the prediction of coronary artery calcification scores.

The implementation of Artificial Intelligence (AI) still faces significant hurdles and one key factor is the access to data. One approach that could support that is federated machine learning (FL) since it allows for privacy preserving data access. For this proof of concept, a prediction model for coronary artery calcification scores (CACS) has been applied. The FL was trained based on the data in the different institutions, while the centralized machine learning model was trained on one allocation of data. Both algorithms predict patients with risk scores ≥5 based on age, biological sex, waist circumference, dyslipidemia and HbA1c. The centralized model yields a sensitivity of c. 66% and a specificity of c. 70%. The FL slightly outperforms that with a sensitivity of 67% while slightly underperforming it with a specificity of 69%. It could be demonstrated that CACS prediction is feasible via both, a centralized and an FL approach, and that both show very comparable accuracy. In order to increase accuracy, additional and a higher volume of patient data is required and for that FL is utterly necessary. The developed "CACulator" serves as proof of concept, is available as research tool and shall support future research to facilitate AI implementation.

Target deconvolution of HDAC pharmacopoeia reveals MBLAC2 as common off-target.

Drugs that target histone deacetylase (HDAC) entered the pharmacopoeia in the 2000s. However, some enigmatic phenotypes suggest off-target engagement. Here, we developed a quantitative chemical proteomics assay using immobilized HDAC inhibitors and mass spectrometry that we deployed to establish the target landscape of 53 drugs. The assay covers 9 of the 11 human zinc-dependent HDACs, questions the reported selectivity of some widely-used molecules (notably for HDAC6) and delineates how the composition of HDAC complexes influences drug potency. Unexpectedly, metallo-ß-lactamase domain-containing protein 2 (MBLAC2) featured as a frequent off-target of hydroxamate drugs. This poorly characterized palmitoyl-CoA hydrolase is inhibited by 24 HDAC inhibitors at low nanomolar potency. MBLAC2 enzymatic inhibition and knockdown led to the accumulation of extracellular vesicles. Given the importance of extracellular vesicle biology in neurological diseases and cancer, this HDAC-independent drug effect may qualify MBLAC2 as a target for drug discovery.

MoSBi: Automated signature mining for molecular stratification and subtyping.

The improving access to increasing amounts of biomedical data provides completely new chances for advanced patient stratification and disease subtyping strategies. This requires computational tools that produce uniformly robust results across highly heterogeneous molecular data. Unsupervised machine learning methodologies are able to discover de novo patterns in such data. Biclustering is especially suited by simultaneously identifying sample groups and corresponding feature sets across heterogeneous omics data. The performance of available biclustering algorithms heavily depends on individual parameterization and varies with their application. Here, we developed MoSBi (molecular signature identification using biclustering), an automated multialgorithm ensemble approach that integrates results utilizing an error model-supported similarity network. We systematically evaluated the performance of 11 available and established biclustering algorithms together with MoSBi. For this, we used transcriptomics, proteomics, and metabolomics data, as well as synthetic datasets covering various data properties. Profiting from multialgorithm integration, MoSBi identified robust group and disease-specific signatures across all scenarios, overcoming single algorithm specificities. Furthermore, we developed a scalable network-based visualization of bicluster communities that supports biological hypothesis generation. MoSBi is available as an R package and web service to make automated biclustering analysis accessible for application in molecular sample stratification.

Kupffer cells are protective in alcoholic steatosis.

Massive accumulation of lipids is a characteristic of alcoholic liver disease. Excess of hepatic fat activates Kupffer cells (KCs), which affect disease progression. Yet, KCs contribute to the resolution and advancement of liver injury. Aim of the present study was to evaluate the effect of KC depletion on markers of liver injury and the hepatic lipidome in liver steatosis (Lieber-DeCarli diet, LDC, female mice, mixed C57BL/6J and DBA/2J background). LDC increased the number of dead hepatocytes without changing the mRNA levels of inflammatory cytokines in the liver. Animals fed LDC accumulated elevated levels of almost all lipid classes. KC ablation normalized phosphatidylcholine and phosphatidylinositol levels in LDC livers, but had no effect in the controls. A modest decline of trigylceride and diglyceride levels upon KC loss was observed in both groups. Serum aminotransferases and hepatic ceramide were elevated in all animals upon KC depletion, and in particular, cytotoxic very long-chain ceramides increased in the LDC livers. Meta-biclustering revealed that eight lipid species occurred in more than 40% of the biclusters, and four of them were very long-chain ceramides. KC loss was further associated with excess free cholesterol levels in LDC livers. Expression of inflammatory cytokines did, however, not increase in parallel. In summary, the current study described a function of KCs in hepatic ceramide and cholesterol metabolism in an animal model of LDC liver steatosis. High abundance of cytotoxic ceramides and free cholesterol predispose the liver to disease progression suggesting a protective role of KCs in alcoholic liver diseases.

Susceptibility to diet-induced obesity at thermoneutral conditions is independent of UCP1.

Activation of uncoupling protein 1 (UCP1) in brown adipose tissue (BAT) upon cold stimulation leads to substantial increase in energy expenditure to defend body temperature. Increases in energy expenditure after a high-caloric food intake, termed diet-induced thermogenesis, are also attributed to BAT. These properties render BAT a potential target to combat diet-induced obesity. However, studies investigating the role of UCP1 to protect against diet-induced obesity are controversial and rely on the phenotyping of a single constitutive UCP1-knockout model. To address this issue, we generated a novel UCP1-knockout model by Cre-mediated deletion of exon 2 in the UCP1 gene. We studied the effect of constitutive UCP1 knockout on metabolism and the development of diet-induced obesity. UCP1 knockout and wild-type mice were housed at 30°C and fed a control diet for 4 wk followed by 8 wk of high-fat diet. Body weight and food intake were monitored continuously over the course of the study, and indirect calorimetry was used to determine energy expenditure during both feeding periods. Based on Western blot analysis, thermal imaging and noradrenaline test, we confirmed the lack of functional UCP1 in knockout mice. However, body weight gain, food intake, and energy expenditure were not affected by loss of UCP1 function during both feeding periods. We introduce a novel UCP1-KO mouse enabling the generation of conditional UCP1-knockout mice to scrutinize the contribution of UCP1 to energy metabolism in different cell types or life stages. Our results demonstrate that UCP1 does not protect against diet-induced obesity at thermoneutrality.NEW & NOTEWORTHY We provide evidence that the abundance of UCP1 does not influence energy metabolism at thermoneutrality studying a novel Cre-mediated UCP1-KO mouse model. This model will be a foundation for a better understanding of the contribution of UCP1 in different cell types or life stages to energy metabolism.

Success Factors of Artificial Intelligence Implementation in Healthcare.

Background: Artificial Intelligence (AI) in healthcare has demonstrated high efficiency in academic research, while only few, and predominantly small, real-world AI applications exist in the preventive, diagnostic and therapeutic contexts. Our identification and analysis of success factors for the implementation of AI aims to close the gap between recent years' significant academic AI advancements and the comparably low level of practical application in healthcare. Methods: A literature and real life cases analysis was conducted in Scopus and OpacPlus as well as the Google advanced search database. The according search queries have been defined based on success factor categories for AI implementation derived from a prior World Health Organization survey about barriers of adoption of Big Data within 125 countries. The eligible publications and real life cases were identified through a catalog of in- and exclusion criteria focused on concrete AI application cases. These were then analyzed to deduct and discuss success factors that facilitate or inhibit a broad-scale implementation of AI in healthcare. Results: The analysis revealed three categories of success factors, namely (1) policy setting, (2) technological implementation, and (3) medical and economic impact measurement. For each of them a set of recommendations has been deducted: First, a risk adjusted policy frame is required that distinguishes between precautionary and permissionless principles, and differentiates among accountability, liability, and culpability. Second, a "privacy by design" centered technology infrastructure shall be applied that enables practical and legally compliant data access. Third, the medical and economic impact need to be quantified, e.g., through the measurement of quality-adjusted life years while applying the CHEERS and PRISMA reporting criteria. Conclusions: Private and public institutions can already today leverage AI implementation based on the identified results and thus drive the translation from scientific development to real world application. Additional success factors could include trust-building measures, data categorization guidelines, and risk level assessments and as the success factors are interlinked, future research should elaborate on their optimal interaction to utilize the full potential of AI in real world application.

Nonalcoholic fatty liver disease stratification by liver lipidomics.

Nonalcoholic fatty liver disease (NAFLD) is a common metabolic dysfunction leading to hepatic steatosis. However, NAFLD's global impact on the liver lipidome is poorly understood. Using high-resolution shotgun mass spectrometry, we quantified the molar abundance of 316 species from 22 major lipid classes in liver biopsies of 365 patients, including nonsteatotic patients with normal or excessive weight, patients diagnosed with NAFL (nonalcoholic fatty liver) or NASH (nonalcoholic steatohepatitis), and patients bearing common mutations of NAFLD-related protein factors. We confirmed the progressive accumulation of di- and triacylglycerols and cholesteryl esters in the liver of NAFL and NASH patients, while the bulk composition of glycerophospho- and sphingolipids remained unchanged. Further stratification by biclustering analysis identified sphingomyelin species comprising n24:2 fatty acid moieties as membrane lipid markers of NAFLD. Normalized relative abundance of sphingomyelins SM 43:3;2 and SM 43:1;2 containing n24:2 and n24:0 fatty acid moieties, respectively, showed opposite trends during NAFLD progression and distinguished NAFL and NASH lipidomes from the lipidome of nonsteatotic livers. Together with several glycerophospholipids containing a C22:6 fatty acid moiety, these lipids serve as markers of early and advanced stages of NAFL.

Investigating Global Lipidome Alterations with the Lipid Network Explorer.

Lipids play an important role in biological systems and have the potential to serve as biomarkers in medical applications. Advances in lipidomics allow identification of hundreds of lipid species from biological samples. However, a systems biological analysis of the lipidome, by incorporating pathway information remains challenging, leaving lipidomics behind compared to other omics disciplines. An especially uncharted territory is the integration of statistical and network-based approaches for studying global lipidome changes. Here we developed the Lipid Network Explorer (LINEX), a web-tool addressing this gap by providing a way to visualize and analyze functional lipid metabolic networks. It utilizes metabolic rules to match biochemically connected lipids on a species level and combine it with a statistical correlation and testing analysis. Researchers can customize the biochemical rules considered, to their tissue or organism specific analysis and easily share them. We demonstrate the benefits of combining network-based analyses with statistics using publicly available lipidomics data sets. LINEX facilitates a biochemical knowledge-based data analysis for lipidomics. It is availableas a web-application and as a publicly available docker container.

Liver Lipids of Patients with Hepatitis B and C and Associated Hepatocellular Carcinoma.

Hepatocellular carcinoma (HCC) still remains a difficult to cure malignancy. In recent years, the focus has shifted to lipid metabolism for the treatment of HCC. Very little is known about hepatitis B virus (HBV) and C virus (HCV)-related hepatic lipid disturbances in non-malignant and cancer tissues. The present study showed that triacylglycerol and cholesterol concentrations were similar in tumor adjacent HBV and HCV liver, and were not induced in the HCC tissues. Higher levels of free cholesterol, polyunsaturated phospholipids and diacylglycerol species were noted in non-tumorous HBV compared to HCV liver. Moreover, polyunsaturated phospholipids and diacylglycerols, and ceramides declined in tumors of HBV infected patients. All of these lipids remained unchanged in HCV-related HCC. In HCV tumors, polyunsaturated phosphatidylinositol levels were even induced. There were no associations of these lipid classes in non-tumor tissues with hepatic inflammation and fibrosis scores. Moreover, these lipids did not correlate with tumor grade or T-stage in HCC tissues. Lipid reprogramming of the three analysed HBV/HCV related tumors mostly resembled HBV-HCC. Indeed, lipid composition of non-tumorous HCV tissue, HCV tumors, HBV tumors and HBV/HCV tumors was highly similar. The tumor suppressor protein p53 regulates lipid metabolism. The p53 and p53S392 protein levels were induced in the tumors of HBV, HCV and double infected patients, and this was significant in HBV infection. Negative correlation of tumor p53 protein with free cholesterol indicates a role of p53 in cholesterol metabolism. In summary, the current study suggests that therapeutic strategies to target lipid metabolism in chronic viral hepatitis and associated cancers have to consider disease etiology.

Loss of hepatic Mboat7 leads to liver fibrosis.

OBJECTIVE: The rs641738C>T variant located near the membrane-bound O-acyltransferase domain containing 7 (MBOAT7) locus is associated with fibrosis in liver diseases, including non-alcoholic fatty liver disease (NAFLD), alcohol-related liver disease, hepatitis B and C. We aim to understand the mechanism by which the rs641738C>T variant contributes to pathogenesis of NAFLD. DESIGN: Mice with hepatocyte-specific deletion of MBOAT7 (Mboat7Δhep) were generated and livers were characterised by histology, flow cytometry, qPCR, RNA sequencing and lipidomics. We analysed the association of rs641738C>T genotype with liver inflammation and fibrosis in 846 NAFLD patients and obtained genotype-specific liver lipidomes from 280 human biopsies. RESULTS: Allelic imbalance analysis of heterozygous human liver samples pointed to lower expression of the MBOAT7 transcript on the rs641738C>T haplotype. Mboat7Δhep mice showed spontaneous steatosis characterised by increased hepatic cholesterol ester content after 10 weeks. After 6 weeks on a high fat, methionine-low, choline-deficient diet, mice developed increased hepatic fibrosis as measured by picrosirius staining (p<0.05), hydroxyproline content (p<0.05) and transcriptomics, while the inflammatory cell populations and inflammatory mediators were minimally affected. In a human biopsied NAFLD cohort, MBOAT7 rs641738C>T was associated with fibrosis (p=0.004) independent of the presence of histological inflammation. Liver lipidomes of Mboat7Δhep mice and human rs641738TT carriers with fibrosis showed increased total lysophosphatidylinositol levels. The altered lysophosphatidylinositol and phosphatidylinositol subspecies in MBOAT7Δhep livers and human rs641738TT carriers were similar. CONCLUSION: Mboat7 deficiency in mice and human points to an inflammation-independent pathway of liver fibrosis that may be mediated by lipid signalling and a potentially targetable treatment option in NAFLD.

Lessons from the COVID-19 pandemic for advancing computational drug repurposing strategies.

Responding quickly to unknown pathogens is crucial to stop uncontrolled spread of diseases that lead to epidemics, such as the novel coronavirus, and to keep protective measures at a level that causes as little social and economic harm as possible. This can be achieved through computational approaches that significantly speed up drug discovery. A powerful approach is to restrict the search to existing drugs through drug repurposing, which can vastly accelerate the usually long approval process. In this Review, we examine a representative set of currently used computational approaches to identify repurposable drugs for COVID-19, as well as their underlying data resources. Furthermore, we compare drug candidates predicted by computational methods to drugs being assessed by clinical trials. Finally, we discuss lessons learned from the reviewed research efforts, including how to successfully connect computational approaches with experimental studies, and propose a unified drug repurposing strategy for better preparedness in the case of future outbreaks.

BALSAM-An Interactive Online Platform for Breath Analysis, Visualization and Classification.

The field of breath analysis lacks a fully automated analysis platform that enforces machine learning good practice and enables clinicians and clinical researchers to rapidly and reproducibly discover metabolite patterns in diseases. We present BALSAM-a comprehensive web-platform to simplify and automate this process, offering features for preprocessing, peak detection, feature extraction, visualization and pattern discovery. Our main focus is on data from multi-capillary-column ion-mobility-spectrometry. While not limited to breath data, BALSAM was developed to increase consistency and robustness in the data analysis process of breath samples, aiming to expand the array of low cost molecular diagnostics in clinics. Our platform is freely available as a web-service and in form of a publicly available docker container.