Showing papers by "Rainer Spang published in 2018"

From hype to reality: data science enabling personalized medicine

Abstract: Personalized, precision, P4, or stratified medicine is understood as a medical approach in which patients are stratified based on their disease subtype, risk, prognosis, or treatment response using specialized diagnostic tests. The key idea is to base medical decisions on individual patient characteristics, including molecular and behavioral biomarkers, rather than on population averages. Personalized medicine is deeply connected to and dependent on data science, specifically machine learning (often named Artificial Intelligence in the mainstream media). While during recent years there has been a lot of enthusiasm about the potential of ‘big data’ and machine learning-based solutions, there exist only few examples that impact current clinical practice. The lack of impact on clinical practice can largely be attributed to insufficient performance of predictive models, difficulties to interpret complex model predictions, and lack of validation via prospective clinical trials that demonstrate a clear benefit compared to the standard of care. In this paper, we review the potential of state-of-the-art data science approaches for personalized medicine, discuss open challenges, and highlight directions that may help to overcome them in the future. There is a need for an interdisciplinary effort, including data scientists, physicians, patient advocates, regulatory agencies, and health insurance organizations. Partially unrealistic expectations and concerns about data science-based solutions need to be better managed. In parallel, computational methods must advance more to provide direct benefit to clinical practice.

TIGIT + iTregs elicited by human regulatory macrophages control T cell immunity

Abstract: Human regulatory macrophages (Mreg) have shown early clinical promise as a cell-based adjunct immunosuppressive therapy in solid organ transplantation. It is hypothesised that recipient CD4+ T cell responses are actively regulated through direct allorecognition of donor-derived Mregs. Here we show that human Mregs convert allogeneic CD4+ T cells to IL-10-producing, TIGIT+ FoxP3+-induced regulatory T cells that non-specifically suppress bystander T cells and inhibit dendritic cell maturation. Differentiation of Mreg-induced Tregs relies on multiple non-redundant mechanisms that are not exclusive to interaction of Mregs and T cells, including signals mediated by indoleamine 2,3-dioxygenase, TGF-β, retinoic acid, Notch and progestagen-associated endometrial protein. Preoperative administration of donor-derived Mregs to living-donor kidney transplant recipients results in an acute increase in circulating TIGIT+ Tregs. These results suggest a feed-forward mechanism by which Mreg treatment promotes allograft acceptance through rapid induction of direct-pathway Tregs.

Correcting for natural isotope abundance and tracer impurity in MS-, MS/MS- and high-resolution-multiple-tracer-data from stable isotope labeling experiments with IsoCorrectoR.

Abstract: Experiments with stable isotope tracers such as 13C and 15N are increasingly used to gain insights into metabolism. However, mass spectrometric measurements of stable isotope labeling experiments should be corrected for the presence of naturally occurring stable isotopes and for impurities of the tracer substrate. Here, we analyzed the effect that such correction has on the data: omitting correction or performing invalid correction can result in largely distorted data, potentially leading to misinterpretation. IsoCorrectoR is the first R-based tool to offer said correction capabilities. It is easy-to-use and comprises all correction features that comparable tools can offer in a single solution: correction of MS and MS/MS data for natural stable isotope abundance and tracer impurity, applicability to any tracer isotope and correction of multiple-tracer data from high-resolution measurements. IsoCorrectoR's correction performance agreed well with manual calculations and other available tools including Python-based IsoCor and Perl-based ICT. IsoCorrectoR can be downloaded as an R-package from: http://bioconductor.org/packages/release/bioc/html/IsoCorrectoR.html .

Cooperative STAT/NF-κB signaling regulates lymphoma metabolic reprogramming and aberrant GOT2 expression.

Abstract: Knowledge of stromal factors that have a role in the transcriptional regulation of metabolic pathways aside from c-Myc is fundamental to improvements in lymphoma therapy. Using a MYC-inducible human B-cell line, we observed the cooperative activation of STAT3 and NF-κB by IL10 and CpG stimulation. We show that IL10 + CpG-mediated cell proliferation of MYClow cells depends on glutaminolysis. By 13C- and 15N-tracing of glutamine metabolism and metabolite rescue experiments, we demonstrate that GOT2 provides aspartate and nucleotides to cells with activated or aberrant Jak/STAT and NF-κB signaling. A model of GOT2 transcriptional regulation is proposed, in which the cooperative phosphorylation of STAT3 and direct joint binding of STAT3 and p65/NF-κB to the proximal GOT2 promoter are important. Furthermore, high aberrant GOT2 expression is prognostic in diffuse large B-cell lymphoma underscoring the current findings and importance of stromal factors in lymphoma biology. Metabolic rewiring of cancer cells can be driven by both extrinsic and intrinsic factors. Here the authors show that microenvironmental factors induce metabolic rewiring of B-cell lymphoma through activation of STAT3 and NF-ΚB resulting in upregulation of the aminotransferase GOT2 and glutamine addiction.

Gene expression profiling reveals a close relationship between follicular lymphoma grade 3A and 3B, but distinct profiles of follicular lymphoma grade 1 and 2

Abstract: A linear progression model of follicular lymphomas (FL) FL1, FL2 and FL3A has been favored, since FL3A often co-exist with an FL1/2 component. FL3B, in contrast, is thought to be more closely related to diffuse large B-cell lymphoma (DLBCL), and both are often simultaneously present in one tumor (DLBCL/FL3B). To obtain more detailed insights into follicular lymphoma progression, a comprehensive analysis of a well-defined set of FL1/2 (n=22), FL3A (n=16), FL3B (n=6), DLBCL/FL3B (n=9), and germinal center B-cell-type diffuse large B-cell lymphoma (n=45) was undertaken using gene expression profiling, immunohistochemical stainings and genetic analyses by fluorescence in situ hybridization. While immunohistochemical (CD10, IRF4/MUM1, Ki67, BCL2, BCL6) and genetic profiles (translocations of BCL2, BCL6 and MYC) delineate FL1-3A from FL3B and DLBCL/FL3B, significant differences were observed between FL1/2 and FL3A upon gene expression profiling. Interestingly, FL3B turned out to be closely related to FL3A, not categorizing within a separate gene expression cluster, and both FL3A and FL3B showed overlapping profiles in between FL1/2 and diffuse large B-cell lymphoma. Finally, based upon their gene expression pattern, DLBCL/FL3B represent a composite form of FL3B and DLBCL, with the majority of samples more closely resembling the latter. The fact that gene expression profiling clearly separated FL1/2 from both FL3A and FL3B suggests a closer biological relationship between the latter. This notion, however, is in contrast to immunohistochemical and genetic profiles of the different histological FL subtypes that point to a closer relationship between FL1/2 and FL3A, and separates them from FL3B.

Advanced patient age at diagnosis of diffuse large B-cell lymphoma is associated with molecular characteristics including ABC-subtype and high expression of MYC

Abstract: The incidence of diffuse large B-cell lymphoma (DLBCL) increases with age being patient age at diagnosis an adverse prognostic factor. However, elderly patients are often underrepresented in common studies. To investigate the effect between age and biological characteristics in DLBCL, we analyzed data of 1534 patients encompassing all adult age groups, enriched for the age ≥75 years. Follicular lymphoma (FL) grade 3B with histopathological characteristics of DLBCLs were included. Gender, centroblastic cytology, FL grade 3B morphology, CD10 expression, and ABC/non-GCB-subtype were significantly associated with age after correction for multiple testing and after adjusting for cohorts. Analysis of a subgroup points towards an association of MYC expression with age. Our data indicate that biological features of DLBCL and FL grade 3B are associated with increasing age among adult patients. The prevalence of the ABC/non-GCB-subtype in elderly patients suggests that therapies targeting this molecular subtype should be specifically explored in this subgroup.

Loss-function learning for digital tissue deconvolution

Abstract: The gene expression profile of a tissue averages the expression profiles of all cells in this tissue. Digital tissue deconvolution (DTD) addresses the following inverse problem: Given the expression profile $y$ of a tissue, what is the cellular composition $c$ of that tissue? If $X$ is a matrix whose columns are reference profiles of individual cell types, the composition $c$ can be computed by minimizing $\mathcal L(y-Xc)$ for a given loss function $\mathcal L$. Current methods use predefined all-purpose loss functions. They successfully quantify the dominating cells of a tissue, while often falling short in detecting small cell populations. Here we learn the loss function $\mathcal L$ along with the composition $c$. This allows us to adapt to application-specific requirements such as focusing on small cell populations or distinguishing phenotypically similar cell populations. Our method quantifies large cell fractions as accurately as existing methods and significantly improves the detection of small cell populations and the distinction of similar cell types.

Round-robin test for the cell-of-origin classification of diffuse large B-cell lymphoma-a feasibility study using full slide staining.

Abstract: Diffuse large B-cell lymphoma (DLBCL) is subdivided by gene expression analysis (GEP) into two molecular subtypes named germinal center B-cell-like (GCB) and activated B-cell-like (ABC) after their putative cell-of-origin (COO). Determination of the COO is considered mandatory in any new-diagnosed DLBCL, not otherwise specified according to the updated WHO classification. Despite the fact that pathologists are free to choose the method for COO classification, immunohistochemical (IHC) assays are most widely used. However, to the best of our knowledge, no round-robin test to evaluate the interlaboratory variability has been published so far. Eight hematopathology laboratories participated in an interlaboratory test for COO classification of 10 DLBCL tumors using the IHC classifier comprising the expression of CD10, BCL6, and MUM1 (so-called Hans classifier). The results were compared with GEP for COO signature and, in a subset, with results obtained by image analysis. In 7/10 cases (70%), at least seven laboratories assigned a given case to the same COO subtype (one center assessed one sample as not analyzable), which was in agreement with the COO subtype determined by GEP. The results in 3/10 cases (30%) revealed discrepancies between centers and/or between IHC and GEP subtype. Whereas the CD10 staining results were highly reproducible, staining for MUM1 was inconsistent in 50% and for BCL6 in 40% of cases. Image analysis of 16 slides stained for BCL6 (N = 8) and MUM1 (N = 8) of the two cases with the highest disagreement in COO classification were in line with the score of the pathologists in 14/16 stainings analyzed (87.5%). This study describes the first round-robin test for COO subtyping in DLBCL using IHC and demonstrates that COO classification using the Hans classifier yields consistent results among experienced hematopathologists, even when variable staining protocols are used. Data from this small feasibility study need to be validated in larger cohorts.

Modelling cancer progression using Mutual Hazard Networks

Abstract: Motivation Cancer progresses by accumulating genomic events, such as mutations and copy number alterations, whose chronological order is key to understanding the disease but difficult to observe. Instead, cancer progression models use co-occurence patterns in cross-sectional data to infer epistatic interactions between events and thereby uncover their most likely order of occurence. State-of-the-art progression models, however, are limited by mathematical tractability and only allow events to interact in directed acyclic graphs, to promote but not inhibit subsequent events, or to be mutually exclusive in distinct groups that cannot overlap. Results Here we propose Mutual Hazard Networks (MHN), a new Machine Learning algorithm to infer cyclic progression models from cross-sectional data. MHN model events by their spontaneous rate of fixation and by multiplicative effects they exert on the rates of successive events. MHN compared favourably to acyclic models in cross-validated model fit on four datasets tested. In application to the glioblastoma dataset from The Cancer Genome Atlas, MHN proposed a novel interaction in line with consecutive biopsies: IDH1 mutations are early events that promote subsequent fixation of TP53 mutations. Availability Implementation and data are available at https://github.com/RudiSchill/MHN.

Loss-function learning for digital tissue deconvolution

Abstract: The gene expression profile of a tissue averages the expression profiles of all cells in this tissue. Digital tissue deconvolution (DTD) addresses the following inverse problem: Given the expression profile y of a tissue, what is the cellular composition c of that tissue? If X is a matrix whose columns are reference profiles of individual cell types, the composition c can be computed by minimizing \(\mathcal {L}(y-Xc)\) for a given loss function \(\mathcal {L}\). Current methods use predefined all-purpose loss functions. They successfully quantify the dominating cells of a tissue, while often falling short in detecting small cell populations.

Fully integrative data analysis of NMR metabolic fingerprints with comprehensive patient data: a case report based on the German Chronic Kidney Disease (GCKD) study

Abstract: Omics data facilitate the gain of novel insights into the pathophysiology of diseases and, consequently, their diagnosis, treatment, and prevention. To that end, it is necessary to integrate omics data with other data types such as clinical, phenotypic, and demographic parameters of categorical or continuous nature. Here, we exemplify this data integration issue for a study on chronic kidney disease (CKD), where complex clinical and demographic parameters were assessed together with one-dimensional (1D) 1H NMR metabolic fingerprints. Routine analysis screens for associations of single metabolic features with clinical parameters, which requires confounding variables typically chosen by expert knowledge to be taken into account. This knowledge can be incomplete or unavailable. The results of this article are manifold. We introduce a framework for data integration that intrinsically adjusts for confounding variables. We give its mathematical and algorithmic foundation, provide a state-of-the-art implementation, and give several sanity checks. In particular, we show that the discovered associations remain significant after variable adjustment based on expert knowledge. In contrast, we illustrate that the discovery of associations in routine analysis can be biased by incorrect or incomplete expert knowledge in univariate screening approaches. Finally, we exemplify how our data integration approach reveals important associations between CKD comorbidities and metabolites. Moreover, we evaluate the predictive performance of the estimated models on independent validation data and contrast the results with a naive screening approach.