Karlin-Altschul statistics

Karlin-Altschul statistics provide a theory for computing the probability that a local alignment of a given score will be found between two random sequences of the same lengths as the query and database sequences. For an introduction, see this page on the NCBI web site . The probability is often expressed as an expectation value , abbreviated to E-value.

K-A statistics apply to local alignments only ; E-values cannot be computed for global alignments.

According to K-A statistics, the expectation value E for a local alignment with score S is:

E = K q d exp(-LS)

Here, q is the query sequence length, d is the database size in letters, exp is the exponential function and K and L are parameters derived from the alignment scoring parameters . L is called the Lambda parameter. USEARCH does not automatically adjust the K and Lambda parameters if the alignment scoring parameters are changed -- they must be provided on the command line.

Option	Default proteins	Default nucleotides	Description
-ka_gapped_lambda	0.267	1.280	Lambda parameter for gapped alignments.
-ka_ungapped_lambda	0.311	1.330	Lambda parameter for ungapped alignments (HSPs).
-ka_gapped_k	0.041	0.460	K parameter for gapped alignments.
-ka_ungapped_k	0.128	0.621	K parameter for ungapped alignments (HSPs).
-ka_dbsize	actual size	actual size	Effective total database size in letters. Most common use is when a database is split into pieces. Then -ka_dbsize should be set to the size of the original database.