Refine
Year of publication
Document Type
- Article (5155)
- Doctoral Thesis (1480)
- Part of Periodical (210)
- Conference Proceeding (164)
- Preprint (164)
- Book (85)
- Contribution to a Periodical (63)
- Review (38)
- Working Paper (22)
- Part of a Book (17)
Language
Keywords
- inflammation (80)
- COVID-19 (58)
- SARS-CoV-2 (48)
- apoptosis (38)
- glioblastoma (38)
- Inflammation (37)
- cancer (37)
- breast cancer (34)
- autophagy (29)
- prostate cancer (28)
Institute
- Medizin (7418) (remove)
Background: Understanding the location and cell-type specific binding of Transcription Factors (TFs) is important in the study of gene regulation. Computational prediction of TF binding sites is challenging, because TFs often bind only to short DNA motifs and cell-type specific co-factors may work together with the same TF to determine binding. Here, we consider the problem of learning a general model for the prediction of TF binding using DNase1-seq data and TF motif description in form of position specific energy matrices (PSEMs).
Methods: We use TF ChIP-seq data as a gold-standard for model training and evaluation. Our contribution is a novel ensemble learning approach using random forest classifiers. In the context of the ENCODE-DREAM in vivo TF binding site prediction challenge we consider different learning setups.
Results: Our results indicate that the ensemble learning approach is able to better generalize across tissues and cell-types compared to individual tissue-specific classifiers or a classifier built based upon data aggregated across tissues. Furthermore, we show that incorporating DNase1-seq peaks is essential to reduce the false positive rate of TF binding predictions compared to considering the raw DNase1 signal.
Conclusions: Analysis of important features reveals that the models preferentially select motifs of other TFs that are close interaction partners in existing protein protein-interaction networks. Code generated in the scope of this project is available on GitHub: https://github.com/SchulzLab/TFAnalysis (DOI: 10.5281/zenodo.1409697).
Background: Understanding the location and cell-type specific binding of Transcription Factors (TFs) is important in the study of gene regulation. Computational prediction of TF binding sites is challenging, because TFs often bind only to short DNA motifs and cell-type specific co-factors may work together with the same TF to determine binding. Here, we consider the problem of learning a general model for the prediction of TF binding using DNase1-seq data and TF motif description in form of position specific energy matrices (PSEMs).
Methods: We use TF ChIP-seq data as a gold-standard for model training and evaluation. Our contribution is a novel ensemble learning approach using random forest classifiers. In the context of the ENCODE-DREAM in vivo TF binding site prediction challenge we consider different learning setups.
Results: Our results indicate that the ensemble learning approach is able to better generalize across tissues and cell-types compared to individual tissue-specific classifiers or a classifier applied to the data aggregated across tissues. Furthermore, we show that incorporating DNase1-seq peaks is essential to reduce the false positive rate of TF binding predictions compared to considering the raw DNase1 signal.
Conclusions: Analysis of important features reveals that the models preferentially select motifs of other TFs that are close interaction partners in existing protein protein-interaction networks. Code generated in the scope of this project is available on GitHub: https://github.com/SchulzLab/TFAnalysis (DOI: 10.5281/zenodo.1409697)
Predicting the requirement for renal replacement therapy in intensive care patients with sepsis
(2018)
Sepsis is one of the most frequent causes of acute kidney injury (AKI) in critically ill patients, with initial organ impairment often followed by dysfunction in other systems. Renal dysfunction may therefore represent one facet in the evolution towards multiple organ dysfunction syndrome (MODS) or, alternatively, may be indicative of system-wide endothelial damage caused by hyperinflammation and a positive fluid balance. Whilst numerous biomarkers have been investigated to predict renal replacement therapy (RRT) requirement, including NGAL, TIMP-2 and IGFBP-7, mid-regional proadrenomedullin (MR-proADM) may also be of interest due to its involvement in capillary leakage, endothelial dysfunction and the initial stages of multiple organ failure development. ...
An expanded chemical space is essential for improved identification of small molecules for emerging therapeutic targets. However, the identification of targets for novel compounds is biased towards the synthesis of known scaffolds that bind familiar protein families, limiting the exploration of chemical space. To change this paradigm, we validated a new pipeline that identifies small molecule-protein interactions and works even for compounds lacking similarity to known drugs. Based on differential mRNA profiles in multiple cell types exposed to drugs and in which gene knockdowns (KD) were conducted, we showed that drugs induce gene regulatory networks that correlate with those produced after silencing protein-coding genes. Next, we applied supervised machine learning to exploit drug-KD signature correlations and enriched our predictions using an orthogonal structure-based screen. As a proof-of-principle for this regimen, top-10/top-100 target prediction accuracies of 26% and 41%, respectively, were achieved on a validation set 152 FDA-approved drugs and 3104 potential targets. We then predicted targets for 1680 compounds and validated chemical interactors with four targets that have proven difficult to chemically modulate, including non-covalent inhibitors of HRAS and KRAS. Importantly, drug-target interactions manifest as gene expression correlations between drug treatment and both target gene KD and KD of genes that act up- or down-stream of the target, even for relatively weak binders. These correlations provide new insights on the cellular response of disrupting protein interactions and highlight the complex genetic phenotypes of drug treatment. With further refinement, our pipeline may accelerate the identification and development of novel chemical classes by screening compound-target interactions.
Simple Summary: The incidence of brain metastases from breast cancer is increasing and the treatment is still a major challenge. Several scores have been developed in order to estimate the prognosis of patients with brain metastases by objective criteria. Here, we validated all three published graded-prognostic-assessment (GPA)-scores in a subcohort of 882 breast cancer patients with brain metastases in the Brain Metastases in the German Breast Cancer (BMBC) registry. Although all three available GPA-scores were associated with OS, they all show limitations mainly in predicting short-term (below 3 months) survival but also in long-term (above 12 months) survival. We discuss the test performances of all scores in our work and provide evidence how physicians should use them as a tool to select patients for different treatment options.
Abstract: Several scores have been developed in order to estimate the prognosis of patients with brain metastases (BM) by objective criteria. The aim of this analysis was to validate all three published graded-prognostic-assessment (GPA)-scores in a subcohort of 882 breast cancer (BC) patients with BM in the Brain Metastases in the German Breast Cancer (BMBC) registry. The median age at diagnosis of BM was 57 years. All in all, 22.3% of patients (n = 197) had triple-negative, 33.4% (n = 295) luminal A like, 25.1% (n = 221) luminal B/HER2-enriched like and 19.2% (n = 169) HER2 positive like BC. Age ≥60 years, evidence of extracranial metastases (ECM), higher number of BM, triple-negative subtype and low Karnofsky-Performance-Status (KPS) were all associated with worse overall survival (OS) in univariate analysis (p < 0.001 each). All three GPA-scores were associated with OS. The breast-GPA showed the highest probability of classifying patients with survival above 12 months in the best prognostic group (specificity 68.7% compared with 48.1% for the updated breast-GPA and 21.8% for the original GPA). Sensitivities for predicting 3 months survival were very low for all scores. In this analysis, all GPA-scores showed only moderate diagnostic accuracy in predicting the OS of BC patients with BM.
Objective To explore factors that potentially impact external validation performance while developing and validating a prognostic model for hospital admissions (HAs) in complex older general practice patients.
Study design and setting Using individual participant data from four cluster-randomised trials conducted in the Netherlands and Germany, we used logistic regression to develop a prognostic model to predict all-cause HAs within a 6-month follow-up period. A stratified intercept was used to account for heterogeneity in baseline risk between the studies. The model was validated both internally and by using internal-external cross-validation (IECV).
Results Prior HAs, physical components of the health-related quality of life comorbidity index, and medication-related variables were used in the final model. While achieving moderate discriminatory performance, internal bootstrap validation revealed a pronounced risk of overfitting. The results of the IECV, in which calibration was highly variable even after accounting for between-study heterogeneity, agreed with this finding. Heterogeneity was equally reflected in differing baseline risk, predictor effects and absolute risk predictions.
Conclusions Predictor effect heterogeneity and differing baseline risk can explain the limited external performance of HA prediction models. With such drivers known, model adjustments in external validation settings (eg, intercept recalibration, complete updating) can be applied more purposefully.
Trial registration number PROSPERO id: CRD42018088129.
Predicting adult Attention Deficit Hyperactivity Disorder (ADHD) using vocal acoustic features
(2021)
Background: It is a key concern in psychiatric research to investigate objective measures to support and ultimately improve diagnostic processes. Current gold standard diagnostic procedures for attention deficit hyperactivity disorder (ADHD) are mainly subjective and prone to bias. Objective measures such as neuropsychological measures and EEG markers show limited specificity. Recent studies point to alterations of voice and speech production to reflect psychiatric symptoms also related to ADHD. However, studies investigating voice in large clinical samples allowing for individual-level prediction of ADHD are lacking. The aim of this study was to explore a role of prosodic voice measures as objective marker of ADHD.
Methods: 1005 recordings were analyzed from 387 ADHD patients, 204 healthy controls, and 100 clinical (psychiatric) controls. All participants (age range 18-59 years, mean age 34.4) underwent an extensive diagnostic examination according to gold standard methods and provided speech samples (3 min in total) including free and given speech. Paralinguistic features were calculated, and random forest based classifications were performed using a 10-fold cross-validation with 100 repetitions controlling for age, sex, and education. Association of voice features and ADHD-symptom severity assessed in the clinical interview were analyzed using random forest regressions.
Results and Conclusion ADHD was predicted with AUC = 0.76. The analysis of a non-comorbid sample of ADHD resulted in similar classification performance. Paralinguistic features were associated with ADHD-symptom severity as indicated by random forest regression. In female participants, particularly with age < 32 years, paralinguistic features showed the highest classification performance (AUC = 0.86).
Paralinguistic features based on derivatives of loudness and fundamental frequency seem to be promising candidates for further research into vocal acoustic biomarkers of ADHD. Given the relatively good performance in female participants independent of comorbidity, vocal measures may evolve as a clinically supportive option in the complex diagnostic process in this patient group.
Competing Interest Statement: EA participated and received payments in the national advisory board ADHD of Shire/Takeda. JL is co-founder and CTO of PeakProfiling GmbH. He created audio-features used in this study, that are intellectual property of PeakProfiling GmbH. FH received payments by PeakProfiling GmbH.
Clinical Trial: NCT01104623
Dysregulation of blood sphingolipids is an emerging topic in clinical science. The objective of this study was to determine preanalytical biases that typically occur in clinical and translational studies and that influence measured blood sphingolipid levels. Therefore, we collected blood samples from four healthy male volunteers to investigate the effect of storage conditions (time, temperature, long-term storage, freeze–thaw cycles), blood drawing (venous or arterial sampling, prolonged venous compression), and sample preparation (centrifugation, freezing) on sphingolipid levels measured by LC-MS/MS. Our data show that sphingosine 1-phosphate (S1P) and sphinganine 1-phosphate (SA1P) were upregulated in whole blood samples in a time- and temperature-dependent manner. Increased centrifugation at higher speeds led to lower amounts of S1P and SA1P. All other preanalytical biases did not significantly alter the amounts of S1P and SA1P. Further, in almost all settings, we did not detect differences in (dihydro)ceramide levels. In summary, besides time-, temperature-, and centrifugation-dependent changes in S1P and SA1P levels, sphingolipids in blood remained stable under practically relevant preanalytical conditions.
Purpose: Optimization of local therapies in synovial sarcoma (SS) considered unresectable at diagnosis is needed. We evaluated the effects of neoadjuvant versus adjuvant radiation versus surgery only on long-term outcomes.
Methods: Patients with macroscopic SS tumors before chemotherapy (IRS-group-III) in the trials CWS-81, CWS-86, CWS-91, CWS-96, CWS-2002-P and SoTiSaR-registry were analyzed. Local therapies were scheduled after 3 neoadjuvant chemotherapy cycles.
Results: Median age of 145 patients was 14.5 years. 106 survivors had median follow-up of 7.0 years. Tumor site was 96 extremities, 19 head–neck, 16 shoulder/hip, 14 trunk. Tumors were < 3 cm in 16, 3–5 cm in 28, 5–10 cm in 55, > 10 cm in 34 patients. In a secondary resection during chemotherapy, R0-status was accomplished in 82, R1 in 30, R2 in 21 (12 missing). Radiotherapy was administered to 115 (R0 61, R1 29, R2 20, missing 5), thereof 57 before and 52 after tumor resection. 23 were treated with surgery only. For all patients, 5 year event-free (EFS) and overall survival (OS) was 68.9% ± 7.6 (95%CI) and 79.1% ± 6.9. To establish independent significance, tumor site, size, surgical results and sequencing of local therapies were analyzed in a Cox regression analysis. Variables associated with EFS and OS are site, size and sequencing of local therapies. Variables associated with local recurrence are site, surgical results and sequencing of local therapies. The only variable associated with suffering metastatic recurrence is tumor size.
Conclusion: Differences in sequencing of local therapy procedures are independently associated with outcomes. Best local control is achieved when tumors are irradiated pre-operatively and undergo R0 or R1 resection thereafter.
Previous magnetoencephalography (MEG) studies have revealed gamma-band activity at sensors over parietal and fronto-temporal cortex during the delay phase of auditory spatial and non-spatial match-to-sample tasks, respectively. While this activity was interpreted as reflecting the memory maintenance of sound features, we noted that task-related activation differences might have been present already prior to the onset of the sample stimulus. The present study focused on the interval between a visual cue indicating which sound feature was to be memorized (lateralization or pitch) and sample sound presentation to test for task-related activation differences preceding stimulus encoding. MEG spectral activity was analyzed with cluster randomization tests (N = 15). Whereas there were no differences in frequencies below 40 Hz, gamma-band spectral amplitude (about 50–65 and 90–100 Hz) was higher for the lateralization than the pitch task. This activity was localized at right posterior and central sensors and present for several hundred ms after task cue offset. Activity at 50–65 Hz was also increased throughout the delay phase for the lateralization compared with the pitch task. Apparently cortical networks related to auditory spatial processing were activated after participants had been informed about the task.