Abstract
Electrophysiological studies indicate altered contrast processing in some Parkinson's Disease (PD) patients. We recently demonstrated that vision is altered in Drosophila PD models and hypothesised that different types of genetic and idiopathic PD may affect dopaminergic visual signalling pathways differently. Here we asked whether visual responses in Drosophila could be used to identify PD mutations. To mimic a clinical setting a range of flies was used. Young flies from four control lines were compared to three early-onset PD mutations (PINK1, DJ-1α and DJ-1β), and to two other neurodegenerative mutations, one in the fly LRRK2 orthologue (dLRRK) the other in eggroll, a model of general neurodegeneration in Drosophila. Stimuli were contrast reversing gratings spanning 64 spatiotemporal frequency combinations. We recorded the steady-state visually-evoked response amplitude across all combinations. We found that the pattern of neuronal responses differed between genotypes. Wild-type and early-onset PD flies formed separate clusters; the late-onset mutation is an outlier. Neuronal responses in early-onset PD flies were stronger than in wild-types. Multivariate pattern analysis grouped flies by PD/non-PD genotype with an accuracy >85%. We propose that machine learning algorithms may be useful in increasing the diagnostic specificity of human electrophysiological measurements in both animal models and PD patients.