Identification of Outlier Loci Responding to Anthropogenic and Natural Selection Pressure in Stream Insects Based on a Self-Organizing Map

Abstract

Water quality maintenance should be considered from an ecological perspective since water is a substrate ingredient in the biogeochemical cycle and is closely linked with ecosystem functioning and services. Addressing the status of live organisms in aquatic ecosystems is a critical issue for appropriate prediction and water quality management. Recently, genetic changes in biological organisms have garnered more attention due to their in-depth expression of environmental stress on aquatic ecosystems in an integrative manner. We demonstrate that genetic diversity would adaptively respond to environmental constraints in this study. We applied a self-organizing map (SOM) to characterize complex Amplified Fragment Length Polymorphisms (AFLP) of aquatic insects in six streams in Japan with natural and anthropogenic variability. After SOM training, the loci compositions of aquatic insects effectively responded to environmental selection pressure. To measure how important the role of loci compositions was in the population division, we altered the AFLP data by flipping the existence of given loci individual by individual. Subsequently we recognized the cluster change of the individuals with altered data using the trained SOM. Based on SOM recognition of these altered data, we determined the outlier loci (over 90th percentile) that showed drastic changes in their belonging clusters (D). Subsequently environmental responsiveness (E k ') was also calculated to address relationships with outliers in different species. Outlier loci were sensitive to slightly polluted conditions including Chl-a, NH 4 -N, NO X -N, PO 4 -P, and SS, and the food material, epilithon. Natural environmental factors such as altitude and sediment additionally showed relationships with outliers in somewhat lower levels. Poly-loci like responsiveness was detected in adapting to environmental constraints. SOM training followed by recognition shed light on developing algorithms de novo to characterize loci information without a priori knowledge of population genetics.

Machine‐learning‐based detection of adaptive divergence of the stream mayfly Ephemera strigata populations

“…As a result, 3 out of the 10 non-neutral loci (loci 56, 89, and 254) showed good model prediction performance (AUC > 0.7), whereas the other 7 non-neutral loci could not be well-modeled. The reason why we could not build a good model for the 7 non-neutral loci is probably because the natural selection on these loci was driven by the other environmental factors which were reported in other studies but not included in our analysis (e.g., velocity, chl-a) (Brouwer, Bessee-Lototskaya, ter Braak, Kraak, & Verdonschot, 2017;Li et al, 2016;Watanabe et al, 2014).…”

Detection of adaptive divergence in populations of the stream mayflyEphemera strigatawith machine learning

“…This may be explained by natural selection at these seven loci being driven by environmental factors not included in our analysis (e.g. velocity and chlorophyll a) (Watanabe et al 2014, Li et al 2016, Brouwer et al 2017). RF is recommended for future studies including huge numbers of environmental variables to assess their effects on adaptive divergence because RF can perform well with large numbers of variables (Genuer et al 2010).…”

Potential Global Invasion Risk of Scale Insect Pests Based on a Self-Organizing Map

“…However, it was difficult to directly detect the boundaries of larger clusters on the trained SOM map. To define the cluster boundaries of the SOM map, a hierarchical cluster analysis with a Ward linkage method [26] was applied [1,11,27]. The Davies-Bouldin index (DBI) [28] was then calculated to select the number of clusters with the minimum DBI value.…”