Synapses are particularly vulnerable in many neurodegenerative diseases and often the first to degenerate, for example in the motor neuron disease spinal muscular atrophy (SMA). Compounds that can counteract synaptic destabilisation are rare. Here, we describe an automated screening paradigm in zebrafish for small-molecule compounds that stabilize the neuromuscular synapse in vivo. We make use of a mutant for the axonal C-type lectin chondrolectin (chodl), one of the main genes dysregulated in SMA. In chodl−/− mutants, neuromuscular synapses that are formed at the first synaptic site by growing axons are not fully mature, causing axons to stall, thereby impeding further axon growth beyond that synaptic site. This makes axon length a convenient read-out for synapse stability. We screened 982 small-molecule compounds in chodl chodl−/− mutants and found four that strongly rescued motor axon length. Aberrant presynaptic neuromuscular synapse morphology was also corrected. The most-effective compound, the adenosine uptake inhibitor drug dipyridamole, also rescued axon growth defects in the UBA1-dependent zebrafish model of SMA. Hence, we describe an automated screening pipeline that can detect compounds with relevance to SMA. This versatile platform can be used for drug and genetic screens, with wider relevance to synapse formation and stabilisation.

Adverse drug events constitute a major challenge for the success of clinical trials. Several computational strategies have been suggested to estimate the risk of adverse drug events in preclinical drug development. While these approaches have demonstrated high utility in practice, they are at the same time limited to specific information sources and thus neglect a wealth of information that is uncovered by fusion of different data sources, including biological protein function, gene expression, chemical compound structure, cell-based imaging, etc. In this work we propose an integrative and explainable Graph Machine Learning approach (MultiGML), which fuses knowledge graphs with multiple further data modalities to predict drug related adverse events. MultiGML demonstrates excellent prediction performance compared to alternative algorithms, including various knowledge graph embedding techniques. MultiGML distinguishes itself from alternative techniques by providing in-depth explanations of model predictions, which point towards biological mechanisms associated with predictions of an adverse drug event.

No abstract

Adverse drug events constitute a major challenge for the success of clinical trials. Several computational strategies have been suggested to estimate the risk of adverse drug events in preclinical drug development. While these approaches have demonstrated high utility in practice, they are at the same time limited to specific information sources and thus neglect a wealth of information that is uncovered by fusion of different data sources, including biological protein function, gene expression, chemical compound structure, cell-based imaging, etc. In this work we propose an integrative and explainable Graph Machine Learning approach (MultiGML), which fuses knowledge graphs with multiple further data modalities to predict drug related adverse events. MultiGML demonstrates excellent prediction performance compared to alternative algorithms, including various knowledge graph embedding techniques. MultiGML distinguishes itself from alternative techniques by providing in-depth explanations of model predictions, which point towards biological mechanisms associated with predictions of an adverse drug event.