Andreas Leha

Recent developments in stem cell biology have enabled the study of cell fate decisions in early human development that are impossible to study in vivo. However, understanding how development varies across individuals and, in particular, the influence of common genetic variants during this process has not been characterised. Here, we exploit human iPS cell lines from 125 donors, a pooled experimental design, and single-cell RNA-sequencing to study population variation of endoderm differentiation. We identify molecular markers that are predictive of differentiation efficiency of individual lines, and utilise heterogeneity in the genetic background across individuals to map hundreds of expression quantitative trait loci that influence expression dynamically during differentiation and across cellular contexts.

CA19-9 values are regularly measured in patients with pancreatic cancer. Certainly, its potential as a biomarker has been compromised by false negative results in CA19-9 negative patients and false positive results in benign pancreatico-biliary diseases. For detection of PDAC recurrence, however, CA19-9 might play an important role. The aim of this study is to analyze the accuracy of CA19-9 for detecting recurrence of pancreatic cancer. All included patients were treated either at the University Medical Center Goettingen, or at the Department of Interdisciplinary Oncology and Pneumonology, DRK-Kliniken Nordhessen, Kassel. We analyzed data of 93 patients with pancreatic cancer in the training set and 41 in the validation set, both retrospectively. Pre- and postoperative CA19-9 values and results of imaging techniques were compared. We performed ROC-analysis. The association between longitudinally measured CA19-9 values and relapse was studied with a joint model between a random effects model for the longitudinal CA19-9 measurements and a Cox proportional hazards models for the survival data. In the test set (n = 93 patients) the median follow-up time was 644 days (22 months). Overall, 71 patients (76.3%) developed recurrence during follow-up. Patients with CA19-9 values of <10kU/l were considered as CA19-9 negative patients (n = 11) and excluded from further analysis. Among the rest, approximately 60% of the patients showed significantly elevated CA19-9 prior to detection of recurrence by imaging techniques. Recurrence was shown by 2.45 times elevated CA19-9 values with 90% positive predictive value. In the validation set, 2.45 times elevated CA19-9 values showed recurrence with 90% sensitivity and 83,33% specificity, with an area under the curve of 95%. Based on measured CA19-9 values during follow-up care, the joint model estimates in recurrence-free patients the probability of recurrence-free survival. CA19-9 elevation is an early and reliable sign for PDAC recurrence. On the strength of a very high accuracy in CA19-9 positive patients, it should be considered to use CA19-9 for therapy decision even without a correlate of imaging technics. Using the joint model, follow-up care of PDAC patients after curative therapy can be stratified.

BACKGROUND: Contemporary deep learning approaches show cutting-edge performance in a variety of complex prediction tasks. Nonetheless, the application of deep learning in healthcare remains limited since deep learning methods are often considered as non-interpretable black-box models. However, the machine learning community made recent elaborations on interpretability methods explaining data point-specific decisions of deep learning techniques. We believe that such explanations can assist the need in personalized precision medicine decisions via explaining patient-specific predictions. METHODS: Layer-wise Relevance Propagation (LRP) is a technique to explain decisions of deep learning methods. It is widely used to interpret Convolutional Neural Networks (CNNs) applied on image data. Recently, CNNs started to extend towards non-Euclidean domains like graphs. Molecular networks are commonly represented as graphs detailing interactions between molecules. Gene expression data can be assigned to the vertices of these graphs. In other words, gene expression data can be structured by utilizing molecular network information as prior knowledge. Graph-CNNs can be applied to structured gene expression data, for example, to predict metastatic events in breast cancer. Therefore, there is a need for explanations showing which part of a molecular network is relevant for predicting an event, e.g., distant metastasis in cancer, for each individual patient. RESULTS: We extended the procedure of LRP to make it available for Graph-CNN and tested its applicability on a large breast cancer dataset. We present Graph Layer-wise Relevance Propagation (GLRP) as a new method to explain the decisions made by Graph-CNNs. We demonstrate a sanity check of the developed GLRP on a hand-written digits dataset and then apply the method on gene expression data. We show that GLRP provides patient-specific molecular subnetworks that largely agree with clinical knowledge and identify common as well as novel, and potentially druggable, drivers of tumor progression. CONCLUSIONS: The developed method could be potentially highly useful on interpreting classification results in the context of different omics data and prior knowledge molecular networks on the individual patient level, as for example in precision medicine approaches or a molecular tumor board.