Noisecut: a python package for noise-tolerant classification of binary data using prior knowledge integration and max-cut solutions.

BACKGROUND: Classification of binary data arises naturally in many clinical applications, such as patient risk stratification through ICD codes. One of the key practical challenges in data classification using machine learning is to avoid overfitting. Overfitting in supervised learning primarily occurs when a model learns random variations from noisy labels in training data rather than the underlying patterns. While traditional methods such as regularization and early stopping have demonstrated effectiveness in interpolation tasks, addressing overfitting in the classification of binary data, in which predictions always amount to extrapolation, demands extrapolation-enhanced strategies. One such approach is hybrid mechanistic/data-driven modeling, which integrates prior knowledge on input features into the learning process, enhancing the model's ability to extrapolate. RESULTS: We present NoiseCut, a Python package for noise-tolerant classification of binary data by employing a hybrid modeling approach that leverages solutions of defined max-cut problems. In a comparative analysis conducted on synthetically generated binary datasets, NoiseCut exhibits better overfitting prevention compared to the early stopping technique employed by different supervised machine learning algorithms. The noise tolerance of NoiseCut stems from a dropout strategy that leverages prior knowledge of input features and is further enhanced by the integration of max-cut problems into the learning process. CONCLUSIONS: NoiseCut is a Python package for the implementation of hybrid modeling for the classification of binary data. It facilitates the integration of mechanistic knowledge on the input features into learning from data in a structured manner and proves to be a valuable classification tool when the available training data is noisy and/or limited in size. This advantage is especially prominent in medical and biomedical applications where data scarcity and noise are common challenges. The codebase, illustrations, and documentation for NoiseCut are accessible for download at https://pypi.org/project/noisecut/ . The implementation detailed in this paper corresponds to the version 0.2.1 release of the software.

Development of a machine learning model for prediction of the duration of unassisted spontaneous breathing in patients during prolonged weaning from mechanical ventilation.

PURPOSE: Treatment of patients undergoing prolonged weaning from mechanical ventilation includes repeated spontaneous breathing trials (SBTs) without respiratory support, whose duration must be balanced critically to prevent over- and underload of respiratory musculature. This study aimed to develop a machine learning model to predict the duration of unassisted spontaneous breathing. MATERIALS AND METHODS: Structured clinical data of patients from a specialized weaning unit were used to develop (1) a classifier model to qualitatively predict an increase of duration, (2) a regressor model to quantitatively predict the precise duration of SBTs on the next day, and (3) the duration difference between the current and following day. 61 features, known to influence weaning, were included into a Histogram-based gradient boosting model. The models were trained and evaluated using separated data sets. RESULTS: 18.948 patient-days from 1018 individual patients were included. The classifier model yielded an ROC-AUC of 0.713. The regressor models displayed a mean absolute error of 2:50 h for prediction of absolute durations and 2:47 h for day-to-day difference. CONCLUSIONS: The developed machine learning model showed informed results when predicting the spontaneous breathing capacity of a patient in prolonged weaning, however lacking prognostic quality required for direct translation to clinical use.

Epigenetic and Transcriptional Shifts in Human Neural Stem Cells after Reprogramming into Induced Pluripotent Stem Cells and Subsequent Redifferentiation.

Induced pluripotent stem cells (iPSCs) and their derivatives have been described to display epigenetic memory of their founder cells, as well as de novo reprogramming-associated alterations. In order to selectively explore changes due to the reprogramming process and not to heterologous somatic memory, we devised a circular reprogramming approach where somatic stem cells are used to generate iPSCs, which are subsequently re-differentiated into their original fate. As somatic founder cells, we employed human embryonic stem cell-derived neural stem cells (NSCs) and compared them to iPSC-derived NSCs derived thereof. Global transcription profiling of this isogenic circular system revealed remarkably similar transcriptomes of both NSC populations, with the exception of 36 transcripts. Amongst these we detected a disproportionately large fraction of X chromosomal genes, all of which were upregulated in iPSC-NSCs. Concurrently, we detected differential methylation of X chromosomal sites spatially coinciding with regions harboring differentially expressed genes. While our data point to a pronounced overall reinstallation of autosomal transcriptomic and methylation signatures when a defined somatic lineage is propagated through pluripotency, they also indicate that X chromosomal genes may partially escape this reinstallation process. Considering the broad application of iPSCs in disease modeling and regenerative approaches, such reprogramming-associated alterations in X chromosomal gene expression and DNA methylation deserve particular attention.

A hybrid modeling framework for generalizable and interpretable predictions of ICU mortality across multiple hospitals.

The development of reliable mortality risk stratification models is an active research area in computational healthcare. Mortality risk stratification provides a standard to assist physicians in evaluating a patient's condition or prognosis objectively. Particular interest lies in methods that are transparent to clinical interpretation and that retain predictive power once validated across diverse datasets they were not trained on. This study addresses the challenge of consolidating numerous ICD codes for predictive modeling of ICU mortality, employing a hybrid modeling approach that integrates mechanistic, clinical knowledge with mathematical and machine learning models . A tree-structured network connecting independent modules that carry clinical meaning is implemented for interpretability. Our training strategy utilizes graph-theoretic methods for data analysis, aiming to identify the functions of individual black-box modules within the tree-structured network by harnessing solutions from specific max-cut problems. The trained model is then validated on external datasets from different hospitals, demonstrating successful generalization capabilities, particularly in binary-feature datasets where label assessment involves extrapolation.

Redefining pandemic preparedness: Multidisciplinary insights from the CERP modelling workshop in infectious diseases, workshop report.

In July 2023, the Center of Excellence in Respiratory Pathogens organized a two-day workshop on infectious diseases modelling and the lessons learnt from the Covid-19 pandemic. This report summarizes the rich discussions that occurred during the workshop. The workshop participants discussed multisource data integration and highlighted the benefits of combining traditional surveillance with more novel data sources like mobility data, social media, and wastewater monitoring. Significant advancements were noted in the development of predictive models, with examples from various countries showcasing the use of machine learning and artificial intelligence in detecting and monitoring disease trends. The role of open collaboration between various stakeholders in modelling was stressed, advocating for the continuation of such partnerships beyond the pandemic. A major gap identified was the absence of a common international framework for data sharing, which is crucial for global pandemic preparedness. Overall, the workshop underscored the need for robust, adaptable modelling frameworks and the integration of different data sources and collaboration across sectors, as key elements in enhancing future pandemic response and preparedness.

Developing an Artificial Intelligence-Based Representation of a Virtual Patient Model for Real-Time Diagnosis of Acute Respiratory Distress Syndrome.

Acute Respiratory Distress Syndrome (ARDS) is a condition that endangers the lives of many Intensive Care Unit patients through gradual reduction of lung function. Due to its heterogeneity, this condition has been difficult to diagnose and treat, although it has been the subject of continuous research, leading to the development of several tools for modeling disease progression on the one hand, and guidelines for diagnosis on the other, mainly the "Berlin Definition". This paper describes the development of a deep learning-based surrogate model of one such tool for modeling ARDS onset in a virtual patient: the Nottingham Physiology Simulator. The model-development process takes advantage of current machine learning and data-analysis techniques, as well as efficient hyperparameter-tuning methods, within a high-performance computing-enabled data science platform. The lightweight models developed through this process present comparable accuracy to the original simulator (per-parameter R2 > 0.90). The experimental process described herein serves as a proof of concept for the rapid development and dissemination of specialised diagnosis support systems based on pre-existing generalised mechanistic models, making use of supercomputing infrastructure for the development and testing processes and supported by open-source software for streamlined implementation in clinical routines.

FOXG1 targets BMP repressors and cell cycle inhibitors in human neural progenitor cells.

FOXG1 is a critical transcription factor in human brain where loss-of-function mutations cause a severe neurodevelopmental disorder, while increased FOXG1 expression is frequently observed in glioblastoma. FOXG1 is an inhibitor of cell patterning and an activator of cell proliferation in chordate model organisms but different mechanisms have been proposed as to how this occurs. To identify genomic targets of FOXG1 in human neural progenitor cells (NPCs), we engineered a cleavable reporter construct in endogenous FOXG1 and performed chromatin immunoprecipitation (ChIP) sequencing. We also performed deep RNA sequencing of NPCs from two females with loss-of-function mutations in FOXG1 and their healthy biological mothers. Integrative analyses of RNA and ChIP sequencing data showed that cell cycle regulation and Bone Morphogenic Protein (BMP) repression gene ontology categories were over-represented as FOXG1 targets. Using engineered brain cell lines, we show that FOXG1 specifically activates SMAD7 and represses CDKN1B. Activation of SMAD7 which inhibits BMP signaling may be one way that FOXG1 patterns the forebrain, while repression of cell cycle regulators such as CDKN1B may be one way that FOXG1 expands the NPC pool to ensure proper brain size. Our data reveal novel mechanisms on how FOXG1 may control forebrain patterning and cell proliferation in human brain development.

Novel architecture for gated recurrent unit autoencoder trained on time series from electronic health records enables detection of ICU patient subgroups.

Electronic health records (EHRs) are used in hospitals to store diagnoses, clinician notes, examinations, lab results, and interventions for each patient. Grouping patients into distinct subsets, for example, via clustering, may enable the discovery of unknown disease patterns or comorbidities, which could eventually lead to better treatment through personalized medicine. Patient data derived from EHRs is heterogeneous and temporally irregular. Therefore, traditional machine learning methods like PCA are ill-suited for analysis of EHR-derived patient data. We propose to address these issues with a new methodology based on training a gated recurrent unit (GRU) autoencoder directly on health record data. Our method learns a low-dimensional feature space by training on patient data time series, where the time of each data point is expressed explicitly. We use positional encodings for time, allowing our model to better handle the temporal irregularity of the data. We apply our method to data from the Medical Information Mart for Intensive Care (MIMIC-III). Using our data-derived feature space, we can cluster patients into groups representing major classes of disease patterns. Additionally, we show that our feature space exhibits a rich substructure at multiple scales.

Analysis of Chest X-ray for COVID-19 Diagnosis as a Use Case for an HPC-Enabled Data Analysis and Machine Learning Platform for Medical Diagnosis Support.

The COVID-19 pandemic shed light on the need for quick diagnosis tools in healthcare, leading to the development of several algorithmic models for disease detection. Though these models are relatively easy to build, their training requires a lot of data, storage, and resources, which may not be available for use by medical institutions or could be beyond the skillset of the people who most need these tools. This paper describes a data analysis and machine learning platform that takes advantage of high-performance computing infrastructure for medical diagnosis support applications. This platform is validated by re-training a previously published deep learning model (COVID-Net) on new data, where it is shown that the performance of the model is improved through large-scale hyperparameter optimisation that uncovered optimal training parameter combinations. The per-class accuracy of the model, especially for COVID-19 and pneumonia, is higher when using the tuned hyperparameters (healthy: 96.5%; pneumonia: 61.5%; COVID-19: 78.9%) as opposed to parameters chosen through traditional methods (healthy: 93.6%; pneumonia: 46.1%; COVID-19: 76.3%). Furthermore, training speed-up analysis shows a major decrease in training time as resources increase, from 207 min using 1 node to 54 min when distributed over 32 nodes, but highlights the presence of a cut-off point where the communication overhead begins to affect performance. The developed platform is intended to provide the medical field with a technical environment for developing novel portable artificial-intelligence-based tools for diagnosis support.

Application of convex hull analysis for the evaluation of data heterogeneity between patient populations of different origin and implications of hospital bias in downstream machine-learning-based data processing: A comparison of 4 critical-care patient datasets.

Machine learning (ML) models are developed on a learning dataset covering only a small part of the data of interest. If model predictions are accurate for the learning dataset but fail for unseen data then generalization error is considered high. This problem manifests itself within all major sub-fields of ML but is especially relevant in medical applications. Clinical data structures, patient cohorts, and clinical protocols may be highly biased among hospitals such that sampling of representative learning datasets to learn ML models remains a challenge. As ML models exhibit poor predictive performance over data ranges sparsely or not covered by the learning dataset, in this study, we propose a novel method to assess their generalization capability among different hospitals based on the convex hull (CH) overlap between multivariate datasets. To reduce dimensionality effects, we used a two-step approach. First, CH analysis was applied to find mean CH coverage between each of the two datasets, resulting in an upper bound of the prediction range. Second, 4 types of ML models were trained to classify the origin of a dataset (i.e., from which hospital) and to estimate differences in datasets with respect to underlying distributions. To demonstrate the applicability of our method, we used 4 critical-care patient datasets from different hospitals in Germany and USA. We estimated the similarity of these populations and investigated whether ML models developed on one dataset can be reliably applied to another one. We show that the strongest drop in performance was associated with the poor intersection of convex hulls in the corresponding hospitals' datasets and with a high performance of ML methods for dataset discrimination. Hence, we suggest the application of our pipeline as a first tool to assess the transferability of trained models. We emphasize that datasets from different hospitals represent heterogeneous data sources, and the transfer from one database to another should be performed with utmost care to avoid implications during real-world applications of the developed models. Further research is needed to develop methods for the adaptation of ML models to new hospitals. In addition, more work should be aimed at the creation of gold-standard datasets that are large and diverse with data from varied application sites.

A training strategy for hybrid models to break the curse of dimensionality.

Mechanistic/data-driven hybrid modeling is a key approach when the mechanistic details of the processes at hand are not sufficiently well understood, but also inferring a model purely from data is too complex. By the integration of first principles into a data-driven approach, hybrid modeling promises a feasible data demand alongside extrapolation. In this work, we introduce a learning strategy for tree-structured hybrid models to perform a binary classification task. Given a set of binary labeled data, the challenge is to use them to develop a model that accurately assesses labels of new unlabeled data. Our strategy employs graph-theoretic methods to analyze the data and deduce a function that maps input features to output labels. Our focus here is on data sets represented by binary features in which the label assessment of unlabeled data points is always extrapolation. Our strategy shows the existence of small sets of data points within given binary data for which knowing the labels allows for extrapolation to the entire valid input space. An implementation of our strategy yields a notable reduction of training-data demand in a binary classification task compared with different supervised machine learning algorithms. As an application, we have fitted a tree-structured hybrid model to the vital status of a cohort of COVID-19 patients requiring intensive-care unit treatment and mechanical ventilation. Our learning strategy yields the existence of patient cohorts for whom knowing the vital status enables extrapolation to the entire valid input space of the developed hybrid model.

Kabuki syndrome stem cell models reveal locus specificity of histone methyltransferase 2D (KMT2D/MLL4).

Kabuki syndrome is frequently caused by loss-of-function mutations in one allele of histone 3 lysine 4 (H3K4) methyltransferase KMT2D and is associated with problems in neurological, immunological and skeletal system development. We generated heterozygous KMT2D knockout and Kabuki patient-derived cell models to investigate the role of reduced dosage of KMT2D in stem cells. We discovered chromosomal locus-specific alterations in gene expression, specifically a 110 Kb region containing Synaptotagmin 3 (SYT3), C-Type Lectin Domain Containing 11A (CLEC11A), Chromosome 19 Open Reading Frame 81 (C19ORF81) and SH3 And Multiple Ankyrin Repeat Domains 1 (SHANK1), suggesting locus-specific targeting of KMT2D. Using whole genome histone methylation mapping, we confirmed locus-specific changes in H3K4 methylation patterning coincident with regional decreases in gene expression in Kabuki cell models. Significantly reduced H3K4 peaks aligned with regions of stem cell maps of H3K27 and H3K4 methylation suggesting KMT2D haploinsufficiency impact bivalent enhancers in stem cells. Preparing the genome for subsequent differentiation cues may be of significant importance for Kabuki-related genes. This work provides a new insight into the mechanism of action of an important gene in bone and brain development and may increase our understanding of a specific function of a human disease-relevant H3K4 methyltransferase family member.

Swarm learning for decentralized artificial intelligence in cancer histopathology.

Artificial intelligence (AI) can predict the presence of molecular alterations directly from routine histopathology slides. However, training robust AI systems requires large datasets for which data collection faces practical, ethical and legal obstacles. These obstacles could be overcome with swarm learning (SL), in which partners jointly train AI models while avoiding data transfer and monopolistic data governance. Here, we demonstrate the successful use of SL in large, multicentric datasets of gigapixel histopathology images from over 5,000 patients. We show that AI models trained using SL can predict BRAF mutational status and microsatellite instability directly from hematoxylin and eosin (H&E)-stained pathology slides of colorectal cancer. We trained AI models on three patient cohorts from Northern Ireland, Germany and the United States, and validated the prediction performance in two independent datasets from the United Kingdom. Our data show that SL-trained AI models outperform most locally trained models, and perform on par with models that are trained on the merged datasets. In addition, we show that SL-based AI models are data efficient. In the future, SL can be used to train distributed AI models for any histopathology image analysis task, eliminating the need for data transfer.

[Usage of Artificial Intelligence in the Combat against the COVID-19 Pandemic]. / Nutzung von künstlicher Intelligenz zur Bekämpfung der COVID-19-Pandemie.

The COVID-19 pandemic is a global health emergency of historic dimension. In this situation, researchers worldwide wanted to help manage the pandemic by using artificial intelligence (AI). This narrative review aims to describe the usage of AI in the combat against COVID-19. The addressed aspects encompass AI algorithms for analysis of thoracic X-rays or CTs, prediction models for severity and outcome of the disease, AI applications in development of new drugs and vaccines as well as forecasting models for spread of the virus. The review shows, which approaches were pursued, and which were successful.

FOXG1 dose tunes cell proliferation dynamics in human forebrain progenitor cells.

Heterozygous loss-of-function mutations in Forkhead box G1 (FOXG1), a uniquely brain-expressed gene, cause microcephaly, seizures, and severe intellectual disability, whereas increased FOXG1 expression is frequently observed in glioblastoma. To investigate the role of FOXG1 in forebrain cell proliferation, we modeled FOXG1 syndrome using cells from three clinically diagnosed cases with two sex-matched healthy parents and one unrelated sex-matched control. Cells with heterozygous FOXG1 loss showed significant reduction in cell proliferation, increased ratio of cells in G0/G1 stage of the cell cycle, and increased frequency of primary cilia. Engineered loss of FOXG1 recapitulated this effect, while isogenic repair of a patient mutation reverted output markers to wild type. An engineered inducible FOXG1 cell line derived from a FOXG1 syndrome case demonstrated that FOXG1 dose-dependently affects all cell proliferation outputs measured. These findings provide strong support for the critical importance of FOXG1 levels in controlling human brain cell growth in health and disease.

Genetic barcoding systematically compares genes in del(5q) MDS and reveals a central role for CSNK1A1 in clonal expansion.

How genetic haploinsufficiency contributes to the clonal dominance of hematopoietic stem cells (HSCs) in del(5q) myelodysplastic syndrome (MDS) remains unresolved. Using a genetic barcoding strategy, we performed a systematic comparison on genes implicated in the pathogenesis of del(5q) MDS in direct competition with each other and wild-type (WT) cells with single-clone resolution. Csnk1a1 haploinsufficient HSCs expanded (oligo)clonally and outcompeted all other tested genes and combinations. Csnk1a1-/+ multipotent progenitors showed a proproliferative gene signature and HSCs showed a downregulation of inflammatory signaling/immune response. In validation experiments, Csnk1a1-/+ HSCs outperformed their WT counterparts under a chronic inflammation stimulus, also known to be caused by neighboring genes on chromosome 5. We therefore propose a crucial role for Csnk1a1 haploinsufficiency in the selective advantage of 5q-HSCs, implemented by creation of a unique competitive advantage through increased HSC self-renewal and proliferation capacity, as well as increased fitness under inflammatory stress.

[Intensive care bed requirements for COVID-19 in the fall/winter of 2021 : Simulation of different scenarios under consideration of incidences and vaccination rates]. / Intensivbettenbedarf für COVID­19 im Herbst/Winter 2021 : Simulation unterschiedlicher Szenarien unter Berücksichtigung von Inzidenzen und Impfquoten.

BACKGROUND: Despite the increasing vaccination rates against SARS-CoV­2, there is a risk of a renewed wave of infections in autumn 2021 due to the high seasonality of the pathogen, with the associated renewed possible heavy burden on intensive care. In the following manuscript we simulated different scenarios using defined mathematical models to estimate the burden of intensive care treatment by COVID-19 patients within certain limits during the coming autumn. METHODS: The simulation of the scenarios uses a stationary model supplemented by the effect of vaccinations. The age group-specific risk profile for intensive care unit (ICU)-associated disease progression is calculated using third wave ICU admission data from sentinel hospitals, local DIVI registry occupancy data and the corresponding local incidence rates by linear regression with time lag. We simulated vaccination rates of 15% for the over 18-year-old cohort, 70% for the 15-34 year cohort, 75%/80%/85% for the 35-59 year cohort and 85%/90%/95% for the over 60-year-old cohort. The simulations take into account that vaccination provides 100% protection against disease progression requiring intensive care. Regarding protection against infection in vaccinated persons the simulations are depicted for the scenario of 70% protection against infection in vaccinated persons and for the scenario of 85% protection against infection in vaccinated persons. RESULTS: The incidence is proportional to ICU bed occupancy. The proportionality factor is higher than in the second and third waves, so that comparable ICU bed occupancy is only achieved at a higher incidence. A 10% increase in vaccination rates of the over 35-year-olds to 85% and of the over 60-year-olds to 95% leads to a significant reduction in ICU bed occupancy. DISCUSSION: There will continue to be a close and linear relationship between SARS-CoV­2 incidence and ICU bed occupancy in the coming months. Even above incidences of 200/100,000 a considerable burden of ICUs with more than 3000 COVID-19 patients can be expected again, unless the vaccination rate is significantly increased. A few percentage points in the vaccination rate have a significant impact on potential ICU occupancy in the autumn, so efforts to increase vaccination acceptance should be a priority in the coming weeks. For intensive care medicine, the vaccination rate of those over 35 years of age is crucial.

Plant PhysioSpace: a robust tool to compare stress response across plant species.

Generalization of transcriptomics results can be achieved by comparison across experiments. This generalization is based on integration of interrelated transcriptomics studies into a compendium. Such a focus on the bigger picture enables both characterizations of the fate of an organism and distinction between generic and specific responses. Numerous methods for analyzing transcriptomics datasets exist. Yet, most of these methods focus on gene-wise dimension reduction to obtain marker genes and gene sets for, for example, pathway analysis. Relying only on isolated biological modules might result in missing important confounders and relevant contexts. We developed a method called Plant PhysioSpace, which enables researchers to compute experimental conditions across species and platforms without a priori reducing the reference information to specific gene sets. Plant PhysioSpace extracts physiologically relevant signatures from a reference dataset (i.e. a collection of public datasets) by integrating and transforming heterogeneous reference gene expression data into a set of physiology-specific patterns. New experimental data can be mapped to these patterns, resulting in similarity scores between the acquired data and the extracted compendium. Because of its robustness against platform bias and noise, Plant PhysioSpace can function as an inter-species or cross-platform similarity measure. We have demonstrated its success in translating stress responses between different species and platforms, including single-cell technologies. We have also implemented two R packages, one software and one data package, and a Shiny web application to facilitate access to our method and precomputed models.

Biometric covariates and outcome in COVID-19 patients: are we looking close enough?

BACKGROUND: The impact of biometric covariates on risk for adverse outcomes of COVID-19 disease was assessed by numerous observational studies on unstratified cohorts, which show great heterogeneity. However, multilevel evaluations to find possible complex, e.g. non-monotonic multi-variate patterns reflecting mutual interference of parameters are missing. We used a more detailed, computational analysis to investigate the influence of biometric differences on mortality and disease evolution among severely ill COVID-19 patients. METHODS: We analyzed a group of COVID-19 patients requiring Intensive care unit (ICU) treatment. For further analysis, the study group was segmented into six subgroups according to Body mass index (BMI) and age. To link the BMI/age derived subgroups with risk factors, we performed an enrichment analysis of diagnostic parameters and comorbidities. To suppress spurious patterns, multiple segmentations were analyzed and integrated into a consensus score for each analysis step. RESULTS: We analyzed 81 COVID-19 patients, of whom 67 required mechanical ventilation (MV). Mean mortality was 35.8%. We found a complex, non-monotonic interaction between age, BMI and mortality. A subcohort of patients with younger age and intermediate BMI exhibited a strongly reduced mortality risk (p < 0.001), while differences in all other groups were not significant. Univariate impacts of BMI or age on mortality were missing. Comparing MV with non-MV patients, we found an enrichment of baseline CRP, PCT and D-Dimers within the MV group, but not when comparing survivors vs. non-survivors within the MV patient group. CONCLUSIONS: The aim of this study was to get a more detailed insight into the influence of biometric covariates on the outcome of COVID-19 patients with high degree of severity. We found that survival in MV is affected by complex interactions of covariates differing to the reported covariates, which are hidden in generic, non-stratified studies on risk factors. Hence, our study suggests that a detailed, multivariate pattern analysis on larger patient cohorts reflecting the specific disease stages might reveal more specific patterns of risk factors supporting individually adapted treatment strategies.