PROSITE documentation PDOC50099

Sequence regions enriched in a particular amino acid profiles

Many proteins contain compositionally biased sequence regions which are also called low-complexity regions [1]. Typically, such regions are highly enriched in one or a few amino acids. We have included profiles specific for each of the 20 amino acids so as to search for regions that are significantly enriched in a particular amino acid (alanine, cysteine, aspartic acid, glutamic acid, phenylalanine, glycine, histidine, isoleucine, lysine, leucine, methionine, asparagine, proline, glutamine, arginine, serine, threonine, valine, tryptophan and tyrosine-rich region profiles). The behaviour of these profiles is controlled by two parameters, the match and mismatch scores. These parameters were chosen such that the "target frequencies" of the corresponding amino acids computed according to the Karlin-Altschul theory [2] approximate 35% for the residue composition of Swiss-Prot (see below).

   Amino     Average    Match    Mismatch   Target
   acid      frequency  score    score      frequency

   Ala (A)   7.55          4       -1       38.5
   Cys (C)   1.69          7       -1       36.8
   Asp (D)   5.30          5       -1       35.1
   Glu (E)   6.32          5       -1       32.4
   Phe (F)   4.07          6       -1       31.9
   Gly (G)   6.84          5       -1       31.2
   His (H)   2.24          7       -1       33.6
   Ile (I)   5.72          5       -1       34.0
   Lys (K)   5.93          5       -1       33.4
   Leu (L)   9.33          4       -1       34.7
   Met (M)   2.35          7       -1       33.1
   Asn (N)   4.52          5       -1       37.4
   Pro (P)   4.92          5       -1       36.2
   Gln (Q)   4.02          6       -1       32.1
   Arg (R)   5.15          5       -1       35.5
   Ser (S)   7.22          4       -1       39.2
   Thr (T)   5.74          5       -1       33.9
   Val (V)   6.52          5       -1       32.0
   Trp (W)   1.25          8       -1       34.9
   Tyr (Y)   3.19          6       -1       35.1

The normalisation parameters for converting raw scores into per-residue log expectation values, which are given within the profile, were empirically derived by fitting an extreme value distribution to the score distribution obtained from a random database that conserves the length distribution and global amino acid composition of Swiss-Prot but not the composition of the individual sequences.