Example:

String 1: ALCALFAAGSKQ
String 2: KCAQFSGGRNPS

Pattern:  CA.F..G

Amino acid classes

these are defined by a mutation probability matrix M and a threshold t. For each amino acid a_i all amino acids a_j such that M(a_i, a_j)>t are considered as an individual class.  For instance, if BLOSUM50 is the mutation probability matrix and t = 1, Then the following amino acid classes are possible:

A  R  N  D  C  Q  E  G  H  I  L  K  M  F  P  S  T  W  Y  V [DE]  [DQE]  [ILMV]  [MIL]  [ST]   [HFWY]  [RK]  [LM]
[FY]  [ST]   [VI]    [ND]   [KQE]  [HY]    [WY] 

Arbitrary matrices are possible to capture any desired clustering scheme. Matrices must be build according to a matrix syntax and must be stored in a matrix subdirectory of the directory where Splash is run.

Mandatory Constraints

The following mandatory constraints must be satisfied by patterns in order for them to be reported:

Minimum Support: There are two choices. 

a) Pattern must occur at least J₀ times in the set of sequences 
b) Pattern must occur in at least J₀ independent sequences

the latter contraint can be expressed as a percentage J₀% of the total number of sequences in the set. 

Density Constraint: Patterns must have at least k₀ matching tokens in each substring of length l₀ that starts with a token. These parameters can be set independently.
Example: pattern CA.F..G satisfies the constraint k₀=2, l₀=4.

Identical matches: Either one or two characters in the patterns must match exactly. This can be set via a command line parameter. This requirement will be removed in a future versions of the program.

Optional Discovery Constraints

Statistical Significance: patterns are reported only if their Z-Score (See Splash paper) exceeds a given z₀ threshold

Length: Patterns are reported only if they have at least l₀ tokens

Identical: Patterns are reported only if all tokens match identically. This is useful in the analysis of DNA or to speed up the analysis in extremely large databases.

Orphan Analysis: Patterns are reported only if they also occur in a predefined sequence. This is useful for orphan analysis.

False positive: Patterns are reported only if the number of expected false positives in SWISS-PROT is below a given threshold.

Output Format Parameters

Format: Patterns can be formatted either as a regular expression, or as PSSM, or as an HTML file. These options are only available with some of the exhaustive or hyerarchical analysis models.

Sort Order: Patterns can be sorted by support (number of occurrences), sequence support (number of distinct sequences in which they occur), length, or ZScore.

Verbose: Patterns can be followed by a list of locations where they occur. These are pairs of Ids and Offsets where they occur starting at Id 1 and Offset 0.