Flowchart of sequence comparison and clustering. Coiled-coil prediction data was generated using the program MultiCoil  and output processing and coiled-coil domain selection were performed as described for the ARABI-COIL database . Coiled-coil prediction data was used to generate a set of sequences with coiled-coil domains masked out. The masked sequences were used as a query against unmasked sequences in an all-against-all Smith-Waterman sequence comparison (SW Search). The P-scores from this comparison were used for clustering of the output.