Graph-theoretical prediction of biological modules in quaternary structures of large protein complexes.

MOTIVATION: The functional complexity of biochemical processes is strongly related to the interplay of proteins and their assembly into protein complexes. In recent years, the discovery and characterization of protein complexes have substantially progressed through advances in cryo-electron microscopy, proteomics, and computational structure prediction. This development results in a strong need for computational approaches to analyse the data of large protein complexes for structural and functional characterization. Here, we aim to provide a suitable approach, which processes the growing number of large protein complexes, to obtain biologically meaningful information on the hierarchical organization of the structures of protein complexes. RESULTS: We modelled the quaternary structure of protein complexes as undirected, labelled graphs called complex graphs. In complex graphs, the vertices represent protein chains and the edges spatial chain-chain contacts. We hypothesized that clusters based on the complex graph correspond to functional biological modules. To compute the clusters, we applied the Leiden clustering algorithm. To evaluate our approach, we chose the human respiratory complex I, which has been extensively investigated and exhibits a known biological module structure experimentally validated. Additionally, we characterized a eukaryotic group II chaperonin TRiC/CCT and the head of the bacteriophage Φ29. The analysis of the protein complexes correlated with experimental findings and indicated known functional, biological modules. Using our approach enables not only to predict functional biological modules in large protein complexes with characteristic features but also to investigate the flexibility of specific regions and coformational changes. The predicted modules can aid in the planning and analysis of experiments. AVAILABILITY AND IMPLEMENTATION: Jupyter notebooks to reproduce the examples are available on our public GitHub repository: https://github.com/MolBIFFM/PTGLtools/tree/main/PTGLmodulePrediction.

Plasma Proteome-Based Test for First-Line Treatment Selection in Metastatic Non-Small Cell Lung Cancer.

PURPOSE: Current guidelines for the management of metastatic non-small cell lung cancer (NSCLC) without driver mutations recommend checkpoint immunotherapy with PD-1/PD-L1 inhibitors, either alone or in combination with chemotherapy. This approach fails to account for individual patient variability and host immune factors and often results in less-than-ideal outcomes. To address the limitations of the current guidelines, we developed and subsequently blindly validated a machine learning algorithm using pretreatment plasma proteomic profiles for personalized treatment decisions. PATIENTS AND METHODS: We conducted a multicenter observational trial (ClinicalTrials.gov identifier: NCT04056247) of patients undergoing PD-1/PD-L1 inhibitor-based therapy (n = 540) and an additional patient cohort receiving chemotherapy (n = 85) who consented to pretreatment plasma and clinical data collection. Plasma proteome profiling was performed using SomaScan Assay v4.1. RESULTS: Our test demonstrates a strong association between model output and clinical benefit (CB) from PD-1/PD-L1 inhibitor-based treatments, evidenced by high concordance between predicted and observed CB (R2 = 0.98, P < .001). The test categorizes patients as either PROphet-positive or PROphet-negative and further stratifies patient outcomes beyond PD-L1 expression levels. The test successfully differentiates between PROphet-negative patients exhibiting high tumor PD-L1 levels (≥50%) who have enhanced overall survival when treated with a combination of immunotherapy and chemotherapy compared with immunotherapy alone (hazard ratio [HR], 0.23 [95% CI, 0.1 to 0.51], P = .0003). By contrast, PROphet-positive patients show comparable outcomes when treated with immunotherapy alone or in combination with chemotherapy (HR, 0.78 [95% CI, 0.42 to 1.44], P = .424). CONCLUSION: Plasma proteome-based testing of individual patients, in combination with standard PD-L1 testing, distinguishes patient subsets with distinct differences in outcomes from PD-1/PD-L1 inhibitor-based therapies. These data suggest that this approach can improve the precision of first-line treatment for metastatic NSCLC.

Does Dual-Energy Computed Tomography Material Decomposition Improve Radiomics Capability to Predict Survival in Head and Neck Squamous Cell Carcinoma Patients? A Preliminary Investigation.

OBJECTIVE: Our study objective was to explore the additional value of dual-energy CT (DECT) material decomposition for squamous cell carcinoma of the head and neck (SCCHN) survival prognostication. METHODS: A group of 50 SCCHN patients (male, 37; female, 13; mean age, 63.6 ± 10.82 years) with baseline head and neck DECT between September 2014 and August 2020 were retrospectively included. Primary tumors were segmented, radiomics features were extracted, and DECT material decomposition was performed. We used independent train and validation datasets with cross-validation and 100 independent iterations to identify prognostic signatures applying elastic net (EN) and random survival forest (RSF). Features were ranked and intercorrelated according to their prognostic importance. We benchmarked the models against clinical parameters. Intraclass correlation coefficients were used to analyze the interreader variation. RESULTS: The exclusively radiomics-trained models achieved similar ( P = 0.947) prognostic performance of area under the curve (AUC) = 0.784 (95% confidence interval [CI], 0.775-0.812) (EN) and AUC = 0.785 (95% CI, 0.759-0.812) (RSF). The additional application of DECT material decomposition did not improve the model's performance (EN, P = 0.594; RSF, P = 0.198). In the clinical benchmark, the top averaged AUC value of 0.643 (95% CI, 0.611-0.675) was inferior to the quantitative imaging-biomarker models ( P < 0.001). A combined imaging and clinical model did not improve the imaging-based models ( P > 0.101). Shape features revealed high prognostic importance. CONCLUSIONS: Radiomics AI applications may be used for SCCHN survival prognostication, but the spectral information of DECT material decomposition did not improve the model's performance in our preliminary investigation.

Dataset of single nucleotide polymorphisms of immune-associated genes in patients with SARS-CoV-2 infection.

The SARS-CoV-2 pandemic has affected nations globally leading to illness, death, and economic downturn. Why disease severity, ranging from no symptoms to the requirement for extracorporeal membrane oxygenation, varies between patients is still incompletely understood. Consequently, we aimed at understanding the impact of genetic factors on disease severity in infection with SARS-CoV-2. Here, we provide data on demographics, ABO blood group, human leukocyte antigen (HLA) type, as well as next-generation sequencing data of genes in the natural killer cell receptor family, the renin-angiotensin-aldosterone and kallikrein-kinin systems and others in 159 patients with SARS-CoV-2 infection, stratified into seven categories of disease severity. We provide single-nucleotide polymorphism (SNP) data on the patients and a protein structural analysis as a case study on a SNP in the SIGLEC7 gene, which was significantly associated with the clinical score. Our data represent a resource for correlation analyses involving genetic factors and disease severity and may help predict outcomes in infections with future SARS-CoV-2 variants and aid vaccine adaptation.

Preliminary results from the EMoLung clinical study showing early lung cancer detection by the LC score.

BACKGROUND: Lung cancer (LC) causes more deaths worldwide than any other cancer type. Despite advances in therapeutic strategies, the fatality rate of LC cases remains high (95%) since the majority of patients are diagnosed at late stages when patient prognosis is poor. Analysis of the International Association for the Study of Lung Cancer (IASLC) database indicates that early diagnosis is significantly associated with favorable outcome. However, since symptoms of LC at early stages are unspecific and resemble those of benign pathologies, current diagnostic approaches are mostly initiated at advanced LC stages. METHODS: We developed a LC diagnosis test based on the analysis of distinct RNA isoforms expressed from the GATA6 and NKX2-1 gene loci, which are detected in exhaled breath condensates (EBCs). Levels of these transcript isoforms in EBCs were combined to calculate a diagnostic score (the LC score). In the present study, we aimed to confirm the applicability of the LC score for the diagnosis of early stage LC under clinical settings. Thus, we evaluated EBCs from patients with early stage, resectable non-small cell lung cancer (NSCLC), who were prospectively enrolled in the EMoLung study at three sites in Germany. RESULTS: LC score-based classification of EBCs confirmed its performance under clinical conditions, achieving a sensitivity of 95.7%, 91.3% and 84.6% for LC detection at stages I, II and III, respectively. CONCLUSIONS: The LC score is an accurate and non-invasive option for early LC diagnosis and a valuable complement to LC screening procedures based on computed tomography.

Variation of butyrate production in the gut microbiome in type 2 diabetes patients / Variación de la producción de butirato en el microbioma intestinal en pacientes con diabetes tipo 2

Background: Diabetes mellitus type 2 is a common disease that poses a challenge to the healthcare system. The disease is very often diagnosed late. A better understanding of the relationship between the gut microbiome and type 2 diabetes can support early detection and form an approach for therapies. Microbiome analysis offers a potential opportunity to find markers for this disease. Next-generation sequencing methods can be used to identify the bacteria present in the stool sample and to generate a microbiome profile through an analysis pipeline. Statistical analysis, e.g., using Student’s t-test, allows the identification of significant differences. The investigations are not only focused on single bacteria, but on the determination of a comprehensive profile. Also, the consideration of the functional microbiome is included in the analyses. The dataset is not from a clinical survey, but very extensive. Results: By examining 946 microbiome profiles of diabetes mellitus type 2 sufferers (272) and healthy control persons (674), a large number of significant genera (25) are revealed. It is possible to identify a large profile for type 2 diabetes disease. Furthermore, it is shown that the diversity of bacteria per taxonomic level in the group of persons with diabetes mellitus type 2 is significantly reduced compared to a healthy control group. In addition, six pathways are determined to be significant for type 2 diabetes describing the fermentation to butyrate. These parameters tend to have high potential for disease detection. Conclusions: With this investigation of the gut microbiome of persons with diabetes type 2 disease, we present significant bacteria and pathways characteristic of this disease.(AU)

Three-dimensional human germinal centers of different sizes in patients diagnosed with lymphadenitis show comparative constant relative volumes of B cells, T cells, follicular dendritic cells, and macrophages.

Germinal centers (GCs) are some of the most important structures in the human immune system. As such, their cell types and functions have been thoroughly investigated. B cells, T cells, follicular dendritic cells (FDCs), and macrophages have widely been found to typically be aggregated in GCs. However, the amount of space occupied by each of these cell types has yet to be investigated. In this study, we conducted confocal laser-based 3D cell-volume quantification of typical GC cells under reactive conditions in lymphadenitis and investigated how volume proportions change during GC development. For this investigation, we used anti-CD3 (T cells), anti-CD20 and anti-Pax5 (B cells), anti-CD23 (FDCs), anti-CD68 (macrophages), and DAPI (nuclear staining). We detected average proportions of about 11% CD3, 9% CD20, 6% CD23, and 2% CD68 in the largest possible regions of interest within GCs. Interestingly, these values remained steady relatively independent of GC size. The remarkably low B cell proportion can be attributed to technical constraints given the use of the CD20 antibody in 3D. Applying the B cell marker Pax5, we found that about 44% of the volume was occupied by B cells after extrapolating the volume of B cell nuclei to that of whole B cells. We concluded that Pax5 is more suitable than anti-CD20 for 3D B cell quantification in GCs. The substantial unstained volume in GCs raises the question of whether other cell types fill these open spaces. Our 3D investigation enabled a unique morphological and volumetric evaluation of GC cells that balance their overall volumes in GCs.

Pathogenetic mechanisms and therapeutic approaches of acute-to-chronic liver failure.

Liver cirrhosis is the end stage of all chronic liver diseases and contributes significantly to overall mortality of 2% globally. The age-standardized mortality from liver cirrhosis in Europe is between 10 and 20% and can be explained by not only the development of liver cancer but also the acute deterioration in the patient's overall condition. The development of complications including accumulation of fluid in the abdomen (ascites), bleeding in the gastrointestinal tract (variceal bleeding), bacterial infections, or a decrease in brain function (hepatic encephalopathy) define an acute decompensation that requires therapy and often leads to acute-on-chronic liver failure (ACLF) by different precipitating events. However, due to its complexity and organ-spanning nature, the pathogenesis of ACLF is poorly understood, and the common underlying mechanisms leading to the development of organ dysfunction or failure in ACLF are still elusive. Apart from general intensive care interventions, there are no specific therapy options for ACLF. Liver transplantation is often not possible in these patients due to contraindications and a lack of prioritization. In this review, we describe the framework of the ACLF-I project consortium funded by the Hessian Ministry of Higher Education, Research and the Arts (HMWK) based on existing findings and will provide answers to these open questions.

Radiomics for therapy-specific head and neck squamous cell carcinoma survival prognostication (part I).

BACKGROUND: Treatment plans for squamous cell carcinoma of the head and neck (SCCHN) are individually decided in tumor board meetings but some treatment decision-steps lack objective prognostic estimates. Our purpose was to explore the potential of radiomics for SCCHN therapy-specific survival prognostication and to increase the models' interpretability by ranking the features based on their predictive importance. METHODS: We included 157 SCCHN patients (male, 119; female, 38; mean age, 64.39 ± 10.71 years) with baseline head and neck CT between 09/2014 and 08/2020 in this retrospective study. Patients were stratified according to their treatment. Using independent training and test datasets with cross-validation and 100 iterations, we identified, ranked and inter-correlated prognostic signatures using elastic net (EN) and random survival forest (RSF). We benchmarked the models against clinical parameters. Inter-reader variation was analyzed using intraclass-correlation coefficients (ICC). RESULTS: EN and RSF achieved top prognostication performances of AUC = 0.795 (95% CI 0.767-0.822) and AUC = 0.811 (95% CI 0.782-0.839). RSF prognostication slightly outperformed the EN for the complete (ΔAUC 0.035, p = 0.002) and radiochemotherapy (ΔAUC 0.092, p < 0.001) cohort. RSF was superior to most clinical benchmarking (p ≤ 0.006). The inter-reader correlation was moderate or high for all features classes (ICC ≥ 0.77 (± 0.19)). Shape features had the highest prognostic importance, followed by texture features. CONCLUSIONS: EN and RSF built on radiomics features may be used for survival prognostication. The prognostically leading features may vary between treatment subgroups. This warrants further validation to potentially aid clinical treatment decision making in the future.

High-resolution transcriptomic and epigenetic profiling identifies novel regulators of COPD.

Patients with chronic obstructive pulmonary disease (COPD) are still waiting for curative treatments. Considering its environmental cause, we hypothesized that COPD will be associated with altered epigenetic signaling in lung cells. We generated genome-wide DNA methylation maps at single CpG resolution of primary human lung fibroblasts (HLFs) across COPD stages. We show that the epigenetic landscape is changed early in COPD, with DNA methylation changes occurring predominantly in regulatory regions. RNA sequencing of matched fibroblasts demonstrated dysregulation of genes involved in proliferation, DNA repair, and extracellular matrix organization. Data integration identified 110 candidate regulators of disease phenotypes that were linked to fibroblast repair processes using phenotypic screens. Our study provides high-resolution multi-omic maps of HLFs across COPD stages. We reveal novel transcriptomic and epigenetic signatures associated with COPD onset and progression and identify new candidate regulators involved in the pathogenesis of chronic lung diseases. The presence of various epigenetic factors among the candidates demonstrates that epigenetic regulation in COPD is an exciting research field that holds promise for novel therapeutic avenues for patients.

The proteasome regulator PSME4 modulates proteasome activity and antigen diversity to abrogate antitumor immunity in NSCLC.

Immunotherapy revolutionized treatment options in cancer, yet the mechanisms underlying resistance in many patients remain poorly understood. Cellular proteasomes have been implicated in modulating antitumor immunity by regulating antigen processing, antigen presentation, inflammatory signaling and immune cell activation. However, whether and how proteasome complex heterogeneity may affect tumor progression and the response to immunotherapy has not been systematically examined. Here, we show that proteasome complex composition varies substantially across cancers and impacts tumor-immune interactions and the tumor microenvironment. Through profiling of the degradation landscape of patient-derived non-small-cell lung carcinoma samples, we find that the proteasome regulator PSME4 is upregulated in tumors, alters proteasome activity, attenuates presented antigenic diversity and associates with lack of response to immunotherapy. Collectively, our approach affords a paradigm by which proteasome composition heterogeneity and function should be examined across cancer types and targeted in the context of precision oncology.

Computational modeling of signal transduction networks without kinetic parameters: Petri net approaches.

More and more computational techniques have been applied to model biological systems, especially signaling pathways in medical systems. Due to the large number of experimental data driven by high-throughput technologies, new computational concepts have been developed. Nevertheless, often the necessary kinetic data cannot be determined in sufficient number and quality because of experimental complexity or ethical reasons. At the same time, the number of qualitative data drastically increased, for example, gene expression data, protein-protein interaction data, and imaging data. Especially for large-scale models, the application of kinetic modeling techniques can fail. On the other hand, many large-scale models have been constructed applying qualitative and semiquantitative techniques, for example, logical models or Petri net models. These techniques make it possible to explore system's dynamics without knowing kinetic parameters. Here, we summarize the work of the last 10 years for modeling signal transduction pathways in medical applications applying Petri net formalism. We focus on analysis techniques based on system's invariants without any kinetic parameters and show predictions of all signaling pathways of the system. We start with an intuitive introduction into Petri nets and system's invariants. We illustrate the main concepts using the tumor necrosis factor receptor 1 (TNFR1)-induced nuclear factor κ-light-chain-enhancer of activated B cells (NF-κB) pathway as a case study. Summarizing recent models, we discuss the advantages and challenges of Petri net applications to medical signaling systems. In addition, we provide exemplarily interesting Petri net applications to model signaling in medical systems of the last years that use the well-known stochastic and kinetic concepts developed about 50 years ago.

USP15 Represses Hepatocellular Carcinoma Progression by Regulation of Pathways of Cell Proliferation and Cell Migration: A System Biology Analysis.

BACKGROUND: Hepatocellular carcinoma (HCC) leads to 600,000 people's deaths every year. The protein ubiquitin carboxyl-terminal hydrolase 15 (USP15) is a ubiquitin-specific protease. The role of USP15 in HCC is still unclear. METHOD: We studied the function of USP15 in HCC from the viewpoint of systems biology and investigated possible implications using experimental methods, such as real-time polymerase chain reaction (qPCR), Western blotting, clustered regularly interspaced short palindromic repeats (CRISPR), and next-generation sequencing (NGS). We investigated tissues samples of 102 patients who underwent liver resection between January 2006 and December 2010 at the Sir Run Run Shaw Hospital (SRRSH). Tissue samples were immunochemically stained; a trained pathologist then scored the tissue by visual inspection, and we compared the survival data of two groups of patients by means of Kaplan-Meier curves. We applied assays for cell migration, cell growth, and wound healing. We studied tumor formation in a mouse model. RESULTS: HCC patients (n = 26) with high expression of USP15 had a higher survival rate than patients (n = 76) with low expression. We confirmed a suppressive role of USP15 in HCC using in vitro and in vivo tests. Based on publicly available data, we constructed a PPI network in which 143 genes were related to USP15 (HCC genes). We combined the 143 HCC genes with results of an experimental investigation to identify 225 pathways that may be related simultaneously to USP15 and HCC (tumor pathways). We found the 225 pathways enriched in the functional groups of cell proliferation and cell migration. The 225 pathways determined six clusters of pathways in which terms such as signal transduction, cell cycle, gene expression, and DNA repair related the expression of USP15 to tumorigenesis. CONCLUSION: USP15 may suppress tumorigenesis of HCC by regulating pathway clusters of signal transduction for gene expression, cell cycle, and DNA repair. For the first time, the tumorigenesis of HCC is studied from the viewpoint of the pathway cluster.

Non-Oncogene Addiction of KRAS-Mutant Cancers to IL-1ß via Versican and Mononuclear IKKß.

Kirsten rat sarcoma virus (KRAS)-mutant cancers are frequent, metastatic, lethal, and largely undruggable. While interleukin (IL)-1ß and nuclear factor (NF)-κB inhibition hold promise against cancer, untargeted treatments are not effective. Here, we show that human KRAS-mutant cancers are addicted to IL-1ß via inflammatory versican signaling to macrophage inhibitor of NF-κB kinase (IKK) ß. Human pan-cancer and experimental NF-κB reporter, transcriptome, and proteome screens reveal that KRAS-mutant tumors trigger macrophage IKKß activation and IL-1ß release via secretory versican. Tumor-specific versican silencing and macrophage-restricted IKKß deletion prevents myeloid NF-κB activation and metastasis. Versican and IKKß are mutually addicted and/or overexpressed in human cancers and possess diagnostic and prognostic power. Non-oncogene KRAS/IL-1ß addiction is abolished by IL-1ß and TLR1/2 inhibition, indicating cardinal and actionable roles for versican and IKKß in metastasis.

Variation of butyrate production in the gut microbiome in type 2 diabetes patients.

BACKGROUND: Diabetes mellitus type 2 is a common disease that poses a challenge to the healthcare system. The disease is very often diagnosed late. A better understanding of the relationship between the gut microbiome and type 2 diabetes can support early detection and form an approach for therapies. Microbiome analysis offers a potential opportunity to find markers for this disease. Next-generation sequencing methods can be used to identify the bacteria present in the stool sample and to generate a microbiome profile through an analysis pipeline. Statistical analysis, e.g., using Student's t-test, allows the identification of significant differences. The investigations are not only focused on single bacteria, but on the determination of a comprehensive profile. Also, the consideration of the functional microbiome is included in the analyses. The dataset is not from a clinical survey, but very extensive. RESULTS: By examining 946 microbiome profiles of diabetes mellitus type 2 sufferers (272) and healthy control persons (674), a large number of significant genera (25) are revealed. It is possible to identify a large profile for type 2 diabetes disease. Furthermore, it is shown that the diversity of bacteria per taxonomic level in the group of persons with diabetes mellitus type 2 is significantly reduced compared to a healthy control group. In addition, six pathways are determined to be significant for type 2 diabetes describing the fermentation to butyrate. These parameters tend to have high potential for disease detection. CONCLUSIONS: With this investigation of the gut microbiome of persons with diabetes type 2 disease, we present significant bacteria and pathways characteristic of this disease.

Holistic View on the Structure of Immune Response: Petri Net Model.

The simulation of immune response is a challenging task because quantitative data are scarce. Quantitative theoretical models either focus on specific cell-cell interactions or have to make assumptions about parameters. The broad variation of, e.g., the dimensions and abundance between lymph nodes as well as between individual patients hampers conclusive quantitative modeling. No theoretical model has been established representing a consensus on the set of major cellular processes involved in the immune response. In this paper, we apply the Petri net formalism to construct a semi-quantitative mathematical model of the lymph nodes. The model covers the major cellular processes of immune response and fulfills the formal requirements of Petri net models. The intention is to develop a model taking into account the viewpoints of experienced pathologists and computer scientists in the field of systems biology. In order to verify formal requirements, we discuss invariant properties and apply the asynchronous firing rule of a place/transition net. Twenty-five transition invariants cover the model, and each is assigned to a functional mode of the immune response. In simulations, the Petri net model describes the dynamic modes of the immune response, its adaption to antigens, and its loss of memory.

Evaluation of automatic discrimination between benign and malignant prostate tissue in the era of high precision digital pathology.

BACKGROUND: Prostate cancer is a major health concern in aging men. Paralleling an aging society, prostate cancer prevalence increases emphasizing the need for efficient diagnostic algorithms. METHODS: Retrospectively, 106 prostate tissue samples from 48 patients (mean age, [Formula: see text] years) were included in the study. Patients suffered from prostate cancer (n = 38) or benign prostatic hyperplasia (n = 10) and were treated with radical prostatectomy or Holmium laser enucleation of the prostate, respectively. We constructed tissue microarrays (TMAs) comprising representative malignant (n = 38) and benign (n = 68) tissue cores. TMAs were processed to histological slides, stained, digitized and assessed for the applicability of machine learning strategies and open-source tools in diagnosis of prostate cancer. We applied the software QuPath to extract features for shape, stain intensity, and texture of TMA cores for three stainings, H&E, ERG, and PIN-4. Three machine learning algorithms, neural network (NN), support vector machines (SVM), and random forest (RF), were trained and cross-validated with 100 Monte Carlo random splits into 70% training set and 30% test set. We determined AUC values for single color channels, with and without optimization of hyperparameters by exhaustive grid search. We applied recursive feature elimination to feature sets of multiple color transforms. RESULTS: Mean AUC was above 0.80. PIN-4 stainings yielded higher AUC than H&E and ERG. For PIN-4 with the color transform saturation, NN, RF, and SVM revealed AUC of [Formula: see text], [Formula: see text], and [Formula: see text], respectively. Optimization of hyperparameters improved the AUC only slightly by 0.01. For H&E, feature selection resulted in no increase of AUC but to an increase of 0.02-0.06 for ERG and PIN-4. CONCLUSIONS: Automated pipelines may be able to discriminate with high accuracy between malignant and benign tissue. We found PIN-4 staining best suited for classification. Further bioinformatic analysis of larger data sets would be crucial to evaluate the reliability of automated classification methods for clinical practice and to evaluate potential discrimination of aggressiveness of cancer to pave the way to automatic precision medicine.

CT-radiomics and clinical risk scores for response and overall survival prognostication in TACE HCC patients.

We aimed to identify hepatocellular carcinoma (HCC) patients who will respond to repetitive transarterial chemoembolization (TACE) to improve the treatment algorithm. Retrospectively, 61 patients (mean age, 65.3 years ± 10.0 [SD]; 49 men) with 94 HCC mRECIST target-lesions who had three consecutive TACE between 01/2012 and 01/2020 were included. Robust and non-redundant radiomics features were extracted from the 24 h post-embolization CT. Five different clinical TACE-scores were assessed. Seven different feature selection methods and machine learning models were used. Radiomics, clinical and combined models were built to predict response to TACE on a lesion-wise and patient-wise level as well as its impact on overall-survival prognostication. 29 target-lesions of 19 patients were evaluated in the test set. Response rates were 37.9% (11/29) on the lesion-level and 42.1% (8/19) on the patient-level. Radiomics top lesion-wise response prognostications was AUC 0.55-0.67. Clinical scores revealed top AUCs of 0.65-0.69. The best working model combined the radiomic feature LargeDependenceHighGrayLevelEmphasis and the clinical score mHAP_II_score_group with AUC = 0.70, accuracy = 0.72. We transferred this model on a patient-level to achieve AUC = 0.62, CI = 0.41-0.83. The two radiomics-clinical features revealed overall-survival prognostication of C-index = 0.67. In conclusion, a random forest model using the radiomic feature LargeDependenceHighGrayLevelEmphasis and the clinical mHAP-II-score-group seems promising for TACE response prognostication.

Comparison of the sensitivity of different criteria to select lung cancer patients for screening in a cohort of German patients.

INTRODUCTION: Trials of CT-based screening for lung cancer have shown a mortality advantage for screening in North America and Europe. Before introducing a nationwide lung cancer screening program in Germany, it is important to assess the criteria used in international trials in the German population. METHODS: We used data from 3623 lung cancer patients from the data warehouse of the German Center for Lung Research (DZL). We compared the sensitivity of the following lung cancer screening criteria overall and stratified by age and histology: the National Lung Screening Trial (NLST), the Danish Lung Cancer Screening Trial (DLCST), the 2013 and 2021 US Preventive Services Task Force (USPSTF), and an adapted version of the Prostate, Lung, Colorectal, and Ovarian no race model (adapted PLCOm2012) with 6-year risk thresholds of 1.0%/6 year and 1.7%/6 year. RESULTS: Overall, the adapted PLCOm2012 model (1%/6 years), selected the highest proportion of lung cancer patients for screening (72.4%), followed by the 2021 USPSTF (70.0%), the adapted PLCOm2012 (1.7%/6 year) (57.4%), the 2013 USPTF (57.0%), DLCST criteria (48.7%), and the NLST (48.5%). The adapted PLCOm2012 risk model (1.0%/6 year) had the highest sensitivity for all histological types except for small-cell and large-cell carcinomas (non-significant), whereas the 2021 USPTF selected a higher proportion of patients. The sensitivity levels were higher in males than in females. CONCLUSION: Using a risk-based selection score resulted in higher sensitivities compared to criteria using dichotomized age and smoking history. However, gender disparities were apparent in all studied eligibility criteria. In light of increasing lung cancer incidences in women, all selection criteria should be reviewed for ways to close this gender gap, especially when implementing a large-scale lung cancer screening program.

Comparison of machine learning methods with logistic regression analysis in creating predictive models for risk of critical in-hospital events in COVID-19 patients on hospital admission.

BACKGROUND: Machine learning (ML) algorithms have been trained to early predict critical in-hospital events from COVID-19 using patient data at admission, but little is known on how their performance compares with each other and/or with statistical logistic regression (LR). This prospective multicentre cohort study compares the performance of a LR and five ML models on the contribution of influencing predictors and predictor-to-event relationships on prediction model´s performance. METHODS: We used 25 baseline variables of 490 COVID-19 patients admitted to 8 hospitals in Germany (March-November 2020) to develop and validate (75/25 random-split) 3 linear (L1 and L2 penalty, elastic net [EN]) and 2 non-linear (support vector machine [SVM] with radial kernel, random forest [RF]) ML approaches for predicting critical events defined by intensive care unit transfer, invasive ventilation and/or death (composite end-point: 181 patients). Models were compared for performance (area-under-the-receiver-operating characteristic-curve [AUC], Brier score) and predictor importance (performance-loss metrics, partial-dependence profiles). RESULTS: Models performed close with a small benefit for LR (utilizing restricted cubic splines for non-linearity) and RF (AUC means: 0.763-0.731 [RF-L1]); Brier scores: 0.184-0.197 [LR-L1]). Top ranked predictor variables (consistently highest importance: C-reactive protein) were largely identical across models, except creatinine, which exhibited marginal (L1, L2, EN, SVM) or high/non-linear effects (LR, RF) on events. CONCLUSIONS: Although the LR and ML models analysed showed no strong differences in performance and the most influencing predictors for COVID-19-related event prediction, our results indicate a predictive benefit from taking account for non-linear predictor-to-event relationships and effects. Future efforts should focus on leveraging data-driven ML technologies from static towards dynamic modelling solutions that continuously learn and adapt to changes in data environments during the evolving pandemic. TRIAL REGISTRATION NUMBER: NCT04659187.