Discovering Novel Signaling Pathways

Problem Statement

Cell signaling research has existed since 1949, with machine learning recently acknowledged as an effective bioinformatics tool. Despite this, only 42 cell signaling papers utilized machine learning methods in 2018. While more than twice the number of papers in 2017 or 2016, no general use tools have been developed and adoption is slow. A primary challenge with machine learning methods in this application is that the input data contains multiple unknown dimensionalities. Machine learning algorithms need to reduce dimensionality before making predictions. The predicted cell signaling pathways are reduced-dimensional data, which needs to be considerably mapped back onto the real world. Biologists need to pay extra attention to how to interpret the prediction. Nowadays, researchers require significant wet-lab experiments to confirm predictions, which requires a large investment of time and financial resources. NicheNet is a recently published tool that could be implemented for bulk and single cell datasets between two datasets/cell types to predict autocrine and paracrine signaling. We plan to add several visualizations on top of NicheNet’s current pipeline to illustrate key signaling interactions between multiple cell types in the cortex and microglia throughout development. Ultimately the goal is to test key predictions about the cell-cell interactions that imprint the tissue specific identity of microglia.

NicheNet Pipeline

Predicted Ligand-Receptor Pairs & Downstream Target Genes

Predicted Signaling Pathway Including Transcriptional Intermediaries in Target Cell