In the field of understanding the relationships between environmental conditions and stream biological assemblages, the aim of the present research was to demonstrate on a pilot data-set the efficiency of using the association of the Self Organizing Map algorithm (Kohonen, 1995) and the Structuring Index (Giraudel, submitted) to classify different ecological conditions and to select the most relevant species (called "structuring species") influencing this typology. Diatoms, water chemistry, stream physical parameters were characterised for 49 reference stations spread over the Adour-Garonne basin (South-Western France). This dataset allowed us to carry on diatom-based stream classification with Self Organizing Maps (SOM), knowing that this methodology has already been successfully used in ecology for community patternising (Foody, 1999 ; Giraudel, 2001 ; Cereghino et al., 2001 ; Brosse et al., 2001), but not yet for benthic algae. The typology obtained was confronted with the corresponding physico-chemical parameters in order to assess relationships between biotic and abiotic descriptors, and then to meet the Water Framework Directive (WFD) requirements concerning stream ecoregional classification. In this way, four main ecoregions were characterised. As SOM enables the observation of each species distribution according to environmental gradients, a specific work has also been led to improve existing ecological profiles (Daget & Godron 1982). This method appeared to be very powerful and promising concerning advances in diatom-based bio-assessment methods. The last part of the work consisted in the selection of relevant diatom species with regards to stream ecotypes. The Structuring Index, a new mathematical tool, allowed us to reach this objective.