Biology subjects

Leha, A.

Publications and source records attributed to Leha, A..

5 recordsLinked to original sources

bioRxiv2021

Stability of feature selection utilizing Graph Convolutional Neural Network and Layer-wise Relevance Propagation

High-throughput technologies are increasingly important in discovering prognostic molecular signatures and identifying novel drug targets. Molecular signatures can be obtained as a subset of features that are important for the decisions of a Machine Learning (ML) method applied to high-dimensional gene expression data. However, feature selection is inherently unstable in this case. Several studies have identified gene sets that provide predictive success for patient prognosis, but these sets usually have only a few genes in common. The stability of feature selection (and reproducibility of identified gene sets) can be improved by including information on molecular networks in ML methods. Graph Convolutional Neural Network (GCNN) is a contemporary deep learning approach applicable to gene expression data structured by a prior knowledge molecular network. Layer-wise Relevance Propagation (LRP) and SHapley Additive exPlanations (SHAP) are techniques to explain individual decisions of deep learning models. We used both GCNN+LRP and GCNN+SHAP techniques to explain GCNNs and to construct feature sets that are relevant to models by aggregating their individual explanations. We also applied more classical ML-based feature selection approaches and analyzed the stability, impact on the classification performance, and interpretability of selected feature sets. Availabilityhttps://gitlab.gwdg.de/UKEBpublic/graph-lrp Contacttim.beissbarth@bioinf.med.uni-goettingen.de

bioinformatics↗

bioRxiv2020

Explaining decisions of Graph Convolutional Neural Networks: patient-specific molecular subnetworks responsible for metastasis prediction in breast cancer

MotivationContemporary 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. 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. ResultsWe 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 applied 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. As a result this method could be potentially highly useful on interpreting classification results on the individual patient level, as for example in precision medicine approaches or a molecular tumor board. Availabilityhttps://gitlab.gwdg.de/UKEBpublic/graph-lrp https://frankkramer-lab.github.io/MetaRelSubNetVis/ Contacttim.beissbarth@bioinf.med.uni-goettingen.de

bioinformatics↗

bioRxiv2020

Iceberg or cut off - how adults who stutter articulate fluent-sounding utterances

Whether fluent-sounding utterances of adults who stutter (AWS) are normally articulated is unclear. We asked 15 AWS and 17 matched adults who do not stutter (ANS) to utter the pseudoword "natscheitideut" 15 times in a 3 T MRI scanner while recording real-time MRI videos at 55 frames per per second in a mid-sagittal plane. All stuttered or otherwise dysfluent runs were discarded. We used sophisticated analyses to model the movement of the tip of the tongue, lips and velum. We observed reproducible movement patterns of the inner and outer articulators which were similar in both groups. Speech duration was similar in both groups and decreased over repetitions, more so in ANS than in AWS. The variability of the movement patterns of tongue, lips and velum decreased over repetitions. The extent of variability decrease was similar in both groups. Across all participants, this repetition effect on movement variability for the lips and the tip of the tongue was less pronounced in severely as compared to mildly stuttering individuals. We conclude that there is no major difference in the movement patterns of a fluent-sounding utterance in both groups. This encourages studies looking at state rather than trait markers of speech dysfluency.

neuroscience↗

bioRxiv2019

Epidermal Growth Factor signaling acts directly and through a sedation neuron to depolarizes a sleep-active neuron following cellular stress

Sleep is induced by sleep-active neurons that depolarize at sleep onset to inhibit wake circuits. Sleep-active neurons are under the control of homeostatic and allostatic mechanisms that determine sleep need. However, little is known about the molecular and circuit mechanisms that translate sleep need into the depolarization of sleep-active neurons. During many conditions in C. elegans sleep induction requires a sleep-active neuron called RIS. Here, we defined the transcriptome of RIS to discover that genes of the Epidermal Growth Factor Receptor (EGFR) signaling pathway are expressed in RIS. With cellular stress, EGFR activates RIS, and RIS induces sleep. Activation of EGFR signaling in the ALA neuron has previously been suggested to promote sleep independently of RIS. Unexpectedly, we found that ALA activation promotes RIS depolarization. Our results suggest that ALA is a sedating neuron with two separable functions. (1) It inhibits specific wakefulness behaviors independently of RIS, (2) and it activates RIS to induce sleep. Whereas ALA plays a strong role in surviving cellular stress, surprisingly, RIS does not. In summary, EGFR signaling can induce sleep-active neuron depolarization by an indirect mechanism through activation of the sedating ALA neuron that acts upstream of the sleep-active RIS neuron as well as through a direct mechanism using EGFR signaling in RIS. Sedation rather than sleep appears to be important for increasing survival following cellular stress, suggesting that sedation and sleep play different roles in restoring health.\n\nHighlights- The transcriptome of the sleep-active RIS neuron reveals the presence of the EGFR signaling machinery\n- EGFR activates RIS directly upon cellular stress to induce sleep bouts\n- In parallel, EGFR activates RIS indirectly through the sedating ALA neuron\n- Sedation rather than sleep bouts support survival following cellular stress

neuroscience↗

bioRxiv2018

Identifying the genetic basis of variation in cell behaviour in human iPS cell lines from healthy donors

Large cohorts of human iPSCs from healthy donors are potentially a powerful tool for investigating the relationship between genetic variants and cellular phenotypes. Here we integrate high content imaging, gene expression and DNA sequence datasets for over 100 human iPSC lines to identify the genetic basis of inter-individual variability in cell behaviour. By applying a dimensionality reduction approach, Probabilistic Estimation of Expression Residuals (PEER), we identified genes that correlated in expression with intrinsic (genetic) and extrinsic (ECM) factors. However, variation in mRNA levels could not account for outlier cell behaviour. Instead, we identified rare, deleterious SNVs in the coding sequence of genes involved in ECM adhesion that occurred in cell lines that were outliers for one or more phenotypes such as cell spreading. These also correlated with altered germ layer differentiation on micropatterned surfaces. Our study thus establishes a strategy for integrating genetic and cell biological measurements for high-throughput analysis.

cell biology↗