Single-cell gene regulatory network prediction by explainable AI

Abstract

The molecular heterogeneity of cancer cells contributes to the often partial response to targeted therapies and relapse of disease due to the escape of resistant cell populations. While single-cell sequencing has started to improve our understanding of this heterogeneity, it offers a mostly descriptive view on cellular types and states. To obtain more functional insights, we propose scGeneRAI, an explainable deep learning approach that uses layer-wise relevance propagation (LRP) to infer gene regulatory networks from static single-cell RNA sequencing data for individual cells. We benchmark our method with synthetic data and apply it to single-cell RNA sequencing data of a cohort of human lung cancers. From the predicted single-cell networks our approach reveals characteristic network patterns for tumor cells and normal epithelial cells and identifies subnetworks that are observed only in (subgroups of) tumor cells of certain patients. While current state-of-the-art methods are limited by their ability to only predict average networks for cell populations, our approach facilitates the reconstruction of networks down to the level of single cells which can be utilized to characterize the heterogeneity of gene regulation within and across tumors.

Molekulare Tumordiagnostik als Triebfeder der Präzisionsonkologie

“…Während Datensätze aus herkömmlichen NGS-Panels, die heutzutage in der Routinediagnostik zur Identifizierung einzelner Mutationen verwendet werden, mit klassischen Datenanalyse-Strategien gehandhabt werden können, profitieren hochdimensionale Genomik-und Proteomik-Datensätze von Verfahren des maschinellen Lernens. Sie helfen dabei die Dimensionalität zu reduzieren und heterogene Datenarten, wie Genregulationen in Netzwerken oder DNA-Methylierungsdaten zu integrieren [49,50,51,52].…”

Decoding pan-cancer treatment outcomes using multimodal real-world data and explainable artificial intelligence

“…We then backpropagate using the generalized LRP-gamma rule, similar to previous works. 43,44 This rule propagates from one layer to the layer below using the equation: where (.) + = max (0, .)…”

“…Here, we used the heuristic 0.01 which worked well in other applications. 44 Applying the rule at each layer, starting at the top layer and moving backwards until the input layer, we obtain in the last step the contribution of each input feature to the prediction. For expanded features, the final LRP score is calculated as the sum of the LRP scores assigned to the tuple (x, 1-x).…”

Precision cancer medicine: Concepts, current practice, and future developments

“…Finally, although image analysis is a major focus of AI developments in pathology as described above, molecular "omics" profiling is another domain with high potential for AI to change the way we deal with such data. Although conventional NGS panels used nowadays in routine diagnostics to identify targetable mutations can be handled with classical data analysis strategies, increasingly common high-dimensional genomics and proteomics may benefit from machine-learning approaches by helping reduce the dimensionality and predict properties such as gene regulatory networks [85,86] or DNA-methylation data [87,88]. In this context, single-cell sequencing approaches are particularly well suited for AI-based analysis as they provide high-dimensional data for a high number of samples (tens of thousands of cells per experiment) [89] and, thus, allow training success beyond what is possible for bulk analyses in which numbers of samples are almost always smaller than the number of molecular measurements.…”