Identification of Synthetic Activators of Cancer Cell Migration by Hybrid Deep Learning.

Deep convolutional neural networks (CNNs) are a method of choice for image recognition. Herein a hybrid CNN approach is presented for molecular pattern recognition in drug discovery. Using self-organizing map images of molecular pharmacophores as input, CNN models were trained to identify chemokine receptor CXCR4 modulators with high accuracy. This machine learning classifier identified first-in-class synthetic CXCR4 full agonists. The receptor-activating effects were confirmed by intracellular cAMP response and in a phenotypic spheroid invasion assay of medulloblastoma cell invasion. Additional macromolecular targets of the small molecules were predicted in silico and tested in vitro, revealing modulatory effects on dopamine receptors and CCR1. These results positively advocate the applicability of molecular image recognition by CNNs to ligand-based virtual compound screening, and demonstrate the complementarity of machine intelligence and human expert knowledge.

Corrigendum: Computer-assisted quantification of motile and invasive capabilities of cancer cells.

This corrects the article DOI: 10.1038/srep15338.

MAP4K4 controlled integrin ß1 activation and c-Met endocytosis are associated with invasive behavior of medulloblastoma cells.

Local tissue infiltration of Medulloblastoma (MB) tumor cells precedes metastatic disease but little is still known about intrinsic regulation of migration and invasion in these cells. We found that MAP4K4, a pro-migratory Ser/Thr kinase, is overexpressed in 30% of primary MB tumors and that increased expression is particularly associated with the frequently metastatic SHH ß subtype. MAP4K4 is a driver of migration and invasion downstream of c-Met, which is transcriptionally up-regulated in SHH MB. Consistently, depletion of MAP4K4 in MB tumor cells restricts HGF-driven matrix invasion in vitro and brain tissue infiltration ex vivo. We show that these pro-migratory functions of MAP4K4 involve the activation of the integrin ß-1 adhesion receptor and are associated with increased endocytic uptake. The consequent enhanced recycling of c-Met caused by MAP4K4 results in the accumulation of activated c-Met in cytosolic vesicles, which is required for sustained signaling and downstream pathway activation. The parallel increase of c-Met and MAP4K4 expression in SHH MB could predict an increased potential of these tumors to infiltrate brain tissue and cause metastatic disease. Molecular targeting of the underlying accelerated endocytosis and receptor recycling could represent a novel approach to block pro-migratory effector functions of MAP4K4 in metastatic cancers.

Computer-assisted quantification of motile and invasive capabilities of cancer cells.

High-throughput analysis of cancer cell dissemination and its control by extrinsic and intrinsic cellular factors is hampered by the lack of adequate and efficient analytical tools for quantifying cell motility. Oncology research would greatly benefit from such a methodology that allows to rapidly determine the motile behaviour of cancer cells under different environmental conditions, including inside three-dimensional matrices. We combined automated microscopy imaging of two- and three-dimensional cell cultures with computational image analysis into a single assay platform for studying cell dissemination in high-throughput. We have validated this new approach for medulloblastoma, a metastatic paediatric brain tumour, in combination with the activation of growth factor signalling pathways with established pro-migratory functions. The platform enabled the detection of primary tumour and patient-derived xenograft cell sensitivity to growth factor-dependent motility and dissemination and identified tumour subgroup-specific responses to selected growth factors of excellent diagnostic value.