Machine Learning models identify gene predictors of waggle dance behaviour in honeybees

Abstract

The molecular characterisation of complex behaviours is a challenging task as a range of different factors are often involved to produce the observed phenotype. An established approach is to look at the overall levels of expression of brain genes – known as ‘neurogenomics’ – to select the best candidates that associate with patterns of interest. This approach has relied so far on a set of powerful statistical tools capable to provide a snapshot of the expression of many thousands of genes that are present in an organism’s genome. However, traditional neurogenomic analyses have some well-known limitations; above all, the limited number of biological replicates compared to the number of genes tested – often referred to as “curse of dimensionality”. Here we implemented a new Machine Learning (ML) approach that can be used as a complement to established methods of transcriptomic analyses. We tested three types of ML models for their performance in the identification of genes associated with honeybee waggle dance. We then intersected the results of these analyses with traditional outputs of differential gene expression analyses and identified two promising candidates for the neural regulation of the waggle dance: the G-protein coupled receptor boss and hnRNP A1, a gene involved in alternative splicing. Overall, our study demonstrates the application of Machine Learning to analyse transcriptomics data and identify genes underlying social behaviour. This approach has great potential for application to a wide range of different scenarios in evolutionary ecology, when investigating the genomic basis for complex phenotypic traits.

Transcriptomic responses to location learning by honeybee dancers are partly mirrored in the brains of dance-followers

“…We repeated the procedure described above on a daily basis, from the end of July to the end of September, alternating the two treatments (vertical stripes and horizontal stripes) for four focal hives, until we had samples for each treatment by colony combination. We used 32 bee samples for the RNAseq experiment, representing the four focal hives across the four experimental groups: eight dancers that perceived long distance + 8 dancers that instead perceived short distance (DDL and DDS groups, respectively); we also added two additional groups composed of bees from the same cohorts as above, hence exposed to the same treatments, that were never seen dancing in the 2 h interval before collection (non-dancers for long and short treatments, NDL and NDS, respectively; see [38] for analyses involving these bees).…”