Conservation Analysis and Secondary Structure Prediction

Conservation analysis was performed with the aid of the AMAS (Analysis of Multiply Aligned Sequences) program [29]. AMAS allows physico chemical properties to be assigned to each amino acid type and the conservation of these properties to be characterised for each position within a sequence alignment. AMAS also allows sub-groups of sequences to be defined so that the similarities and differences between sub-families can be rapidly discerned.

The turn prediction algorithms of Wilmot and Thornton [30] and Rose [31] were applied to each sequence in the alignment. Predictions of - helix and - strand were performed using the Robson [32], Lim [33] and Chou &Fasman [34] methods. For sequences this gives predictions for turn/loop, predictions for - helix and predictions for - strand at each aligned position. The total possible number of positive turn predictions at a position is , similarly, the total possible - helix or - strand prediction at a position is . A ``consensus'' secondary structure prediction was obtained by dividing the total number of positive predictions for each state at each position by these maxima. A preliminary secondary structure for the position was then assigned by taking the state with the highest fraction. For example, if we have 10 sequences, at each position there will be 20 possible turn predictions, 30 possible helix predictions and 30 strand predictions. If there are actually 15, 15, 20, then the fractions are turn:15/20 = 0.75, helix: 15/30 = 0.5, strand: 20/30 = 0.67; the position would therefore be assigned to turn. This preliminary prediction was then interpreted in the light of the conservation patterns seen across the complete alignment (Figure 1). All alignment figures were prepared using the ALSCRIPT program [35].