Welcome to the TaxSOM Web interface!

TaxSOM is a tool for taxonomic classification of DNA fragments, as they are typically obtained in metagenome projects. The classification is based on taxon-specific DNA base composition characteristics (genomic signatures). Classification of query sequences is achieved by mapping the query sequences to the genomic signatures of sequences with known taxonomic affiliations. The mapping is done via the Self-Organizing Map (SOM) algorithm.

TaxSOM offers two modes of operation:

  • mapping query sequences to a pre-calculated SOM
  • mapping query sequences to a custom SOM

The basic mode of TaxSOM is to map query sequences to a pre-calculated SOM. This involves the following steps:

  • upload of query sequences (currently max. 50 MB)
  • SOM selection
  • calculation start
  • visual inspection and/or download of the results

The more advanced mode of TaxSOM involves the construction of a custom SOM with a subsequent mapping of the query sequences. This involves the following steps:

  • selection of training sequences for SOM calculation. Sequence selection can be done via
    • the TaxSOM taxonomy browser
    • upload of sequences
  • selection of parameters for SOM construction
  • start of SOM calculation
  • the remaining steps are the same as for the mapping to a pre-calculated SOM

Under the hood

Two SOM variants are implemented in TaxSOM, the batch-learning SOM (BLSOM) and the growing SOM (GSOM). Both SOM variants can be fed with either raw oligonucleotide counts (better for sequences < 5 kb) or with maximum Markov model oligonucleotide frequency statistics ( better for sequences > 5 kb).
As a starting point, the following articles may be of interest:

Informatics for Unveiling Hidden Genome Signatures Takashi Abe, et. al. Genome Res. 2003 April 1; 13(4): 693-702

Dynamic Self Organising Maps with Controlled Growth for Knowledge Discovery, Alahakoon D., et. al. IEEE Transactions on Neural Networks, Volume 11, No. 3, pp 601-614, May 2000


If you use this tool, please cite.
Weber, M., et al.: Practical Application of Self-Organizing Maps to interrelate biodiversity and functional data in NGS-based environmental metagenomics

  • Select training-data from the NCBI taxonomy

  • Upload own data for training and/or classification

  • View calculated SOMs via SVG-graphics

  • Classification-results displayed as table or csv-file