Figure 1 . Schematic Representation of the Experimental Design. A) Honeybees visited a feeder through a tunnel, used to alter their distance perception (vertical stripes with respect to the direction of flight increased, while horizontal stripes decreased the distance perceived), and were then marked accordingly. B) At the observation hive, honeybees were recorded while performing or not-performing dancing behaviour and were finally prepared for RNAseq analysis (C). D) Three machine learning algorithms (SVM, GLMNET and RFE) were trained on the pre-processed sequence reads (Training Classifiers). E) Key features from each model were compared to identify common elements (genes or predictors).
Table 1 . Benchmark Algorithms. We chose to test SVM, GLMNET, RF and RFE for our study, based on their use in previous research. The first three algorithms use embedded feature selection (FS) to obtain key predictors from the trained model (Embedded), while RFE requires an underlying embedded approach for the ranking (Wrapper). We report the studies that featured or reviewed these algorithms.