MatDefine: Definition of weight matrices

[Introduction] [Sequence Selection] [Parameters] [Output] [References]

Introduction

MatDefine is a tool for fully automatic definition and evaluation of weight matrices from a set of short DNA sequences. The resulting weight matrix can be used by MatInspector to scan nucleic acid sequences for matches to the described binding site.

The quality of a matrix is estimated by a value for random expectation (RE-value), which is defined as the number of matches with high matrix similarity (>= 0.85) expected in a random sequence of 1000 bp. This RE-value is assigned to each matrix.

Per default, the weight matrix is generated without any user interaction. A protocol describing the matrix definition process is delivered. The following steps are performed:

definition of an optimum common core sequence
core-anchored alignment of all sequences
generation of a weight matrix consistent with the training set
strand optimization of all sequences (if selected)
determination of length of weight matrix
determination of optimized threshold (for transcription factor binding site matrices)

All default parameters can be changed. The following additional options are available:

unanchored alignment of sequences
QuickAlign (alignment without introducing gaps)
remove identical sequences
check all sequences against MatInspector library

In case of unsuitable input sequences, no matrix will be generated.

Input Data

Sequence Selection
Input data	There are several ways to supply the input data for MatDefine: Choose from your previously uploaded sequences. Enter the input data directly into the form. If your browser supports this option, an input file can be uploaded. In all cases two different types of input data are accepted: sequences or nucleotide distribution matrices. Sequences The input sequences have to be in either EMBL format, GenBank format, FASTA format, or IG format The sequence file may contain at most 5000 sequences. Nucleotide distribution matrix The nucleotide distribution matrix file consists of five lines. The first line has to contain the numbers of A at each position of the binding site, the 2nd, 3rd, and 4th line the numbers of C, G, T, respectively. The fifth line has to contain the number of elements (sequences) from which the matrix was derived. The matrix elements can be separated by any character (except '.'). The following is an example of a nucleotide distribution matrix file: 4 0 0 0 1 7 0 22 0 0 0 0 7 5 4 3 4 5 4 3 9 8 2 22 0 0 0 22 15 5 4 7 4 4 2 4 4 2 1 7 0 0 0 0 0 0 2 7 3 6 3 2 7 9 10 12 6 0 0 22 22 0 7 8 6 8 5 1 22 elements In this example matrix elements are separated by blanks.
Parameters
Note: The following parameters are only relevant for sequence input files. If your input is a nucleotide distribution matrix these parameters will be ignored.
Strand optimization	The strand optimization option is useful if the orientation of the binding site is unknown. In this case both strands of the input sequences are checked and the "+" or "-" strand is used for matrix definition. If this option is selected, core-anchored alignment is used automatically.
Alignment and Tuple Search	Core-anchored alignment With core-anchored alignment, the best conserved core-tuple is selected and the alignment is anchored at the first position of the core-tuple in each sequence. The tuple selection algorithm is described in the CoreSearch paper. The following parameters can be modified: Length of core-tuple: The minimum and optimum length of the core-tuple to be searched in the input sequences is 4 bp. If a higher tuple length is selected and no common tuple can be found, smaller tuples are searched until a common tuple can be identified or the minimum tuple length is reached. If matrix generation fails using the input tuple length, the tuple length is increased up to the maximum tuple length of 8 bp until a matrix can be created. The core-tuple needs not to be identical in all sequences, it may contain a single mismatch at a random position. Minimum number of sequences containing tuple: This is the minimum percent of sequences in which the core-tuple has to be present. Sequences for which no core-tuple can be found will be rejected. QuickAlign: QuickAlign is an alignment without introducing gaps. It is about three times faster as the default alignment. QuickAlign is used automatically if the input consists of more than 100 sequences. These parameters are hidden by default. You can use the next to the section header to reveal them!
Alignment and Tuple Search	Unanchored alignment Unanchored alignment means that the search for a common core sequence will be omitted. You should use this option if your sequences do not contain a highly conserved core sequence. These parameters are hidden by default. You can use the next to the section header to reveal them!
Matrix Creation	Cut off matrix ends MatDefine automatically determines the correct length of the matrix by cutting off low conserved positions at both matrix ends. For example, if the input sequences are very different in length or contain sequences around the binding site it is necessary to reduce the matrix length. Turning off this feature makes only sense if you are sure that your input sequences are confined to the binding sites. These parameters are hidden by default. You can use the next to the section header to reveal them!
	Remove identical sequences Identical sequences (i.e. one sequence equals another sequence or is part of another sequence) can be removed to avoid a biased nucleotide distribution matrix. In case of sequences with different length the shorter sequence will be removed. Regardless of this option, MatDefine always identifies identical sequences in the output file. These parameters are hidden by default. You can use the next to the section header to reveal them!
	Calculate optimized threshold The optimized threshold of a weight matrix is the matrix similarity threshold that minimizes false positive matches when the matrix is used to scan sequences with MatInspector. It is defined in a way that at most 3 matches are found in 10,000 bp of non-regulatory test sequences (i.e. with the optimized threshold less than 3 false positives per 10,000 bp are found). Since the calculation of the optimized threshold requires some computing time it can be omitted for test runs. These parameters are hidden by default. You can use the next to the section header to reveal them!
Consistency Check	Minimum number of sequences This is the minimum number of sequences which is required to define a matrix. If the input file contains less sequences or less sequences remain after the rejection process no matrix will be created. These parameters are hidden by default. You can use the next to the section header to reveal them!
Consistency Check	Minimum matrix similarity MatDefine generates a weight matrix which is consistent with its training set, i.e. all training sequences have to be identified by the resulting matrix. Therefore, only sequences that reach the minimum matrix similarity are included in the matrix, all other sequences are rejected. Decreasing the minimum matrix similarity may lead to inclusion of more sequences but also can influence the quality of the matrix. Increasing the minimum matrix similarity may lead to rejection of more sequences. If too few sequences are retained, no matrix will be created (see minimum number of sequences). These parameters are hidden by default. You can use the next to the section header to reveal them!
Library Comparison	Selection of matrix groups Here, you can select the matrix groups from the current MatInspector library with which the newly generated matrix should be compared. Per default, all matrix groups including your user-defined matrices (if available) are selected. Please note that the library comparison cannot be completely disabled. If you do not select at least one matrix group, the new matrix will be compared with all available matrix groups. These parameters are hidden by default. You can use the next to the section header to reveal them!
Library Comparison	Check all sequences If this option is set, all input sequences will be checked against the matrix groups selected above (using optimized matrix similarity threshold). For each sequence you receive the list of matrix families with at least one match in this sequence. These parameters are hidden by default. You can use the next to the section header to reveal them!
Your email address	Here you can choose between two methods for receiving the results: Show result directly in browser window In this option the URL of the result is directly shown in your browser window. Warning: Please use this option only for analyses which can be performed in a short time. If the analysis takes longer than the timeout of the webserver, the connection will be terminated and you will receive an error message (e.g. "The document contained no data."). In this case, the results will not be available, please restart the analysis using the option below "Send the URL of the result to". Send the URL of the result via email In this option an email with the URL of the results will be sent to the user provided email address, when the analysis is finished. The results will be available for a limited time on our server. For details of how long your results will be kept please see the result-email. After that period they will be deleted unless protected in the project management!

Program Output

MatDefine creates a protocol detailing each step of the matrix generation, and the weight matrix which is used by MatInspector. The resulting matrix can be saved to your personal matrix library (user-defined library).

The protocol file contains

the alignment of the selected sequences
the matrix similarity of each selected sequence
the list of identical sequences
the list of rejected sequences including the reason for rejection
the matrix families of the current MatInspector library which are similar to the new matrix

Example Protocol Output

Identical sequences

Sequence	Identical to
HSFOS	MMCFOS
XLACTIN5A	XLACTIN8A

Alignment

Core sequence:	CCAT
Number of aligned sequences:	20
Number of rejected sequences:	0

Sequence Name	Position	Str.	Alignment	Matrix Similarity
MMTFEZIF2 MMTFEZIF1 HSACTCA2 XLACTCAG3 EBV GGACAREG1 GGACAREG2 HSACTBPR HSVLC1 MMCYR61G XLACTIN8A XLACTIN5A XLACTCAG1 HSACTCA3 HSACTCA4 XLACTCAG2 MMCFOS HSFOS MMKROX1 MMTFEZIF	4 - 23 4 - 23 4 - 23 4 - 23 4 - 23 4 - 23 3 - 22 3 - 22 4 - 23 4 - 23 4 - 23 4 - 23 3 - 22 3 - 22 3 - 22 3 - 22 3 - 22 3 - 22 3 - 22 3 - 22	(+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+) (+)	CG CCAT ATAAGGAGCAGGAA CG CCTT ATATGGAGTGGCCC GA CCAA ATAAGGCAAGGTGG TA CCAA ATAAGGGCAGGCTG AG CCAT ATGTGGACAGATGG CG CCTT CTTTGGGCAGCGCG AC CCAA ATATGGCGACGGCC GT CCTT ATATGGACTCATCT AT CCTT TTATGGCCCTGTCC AC CCAA ATATGGAAATATTG GC CCAT ATTTGGCGATCTTC GC CCAT ATTTGGCGATCTTC AT CCCT ATTTGGCCATCCCT CT CCCT ATTTGGCCATCCCC TT CCTT ACATGGTCTGGGGG TT CCAT ACATGGGCTAAGGG GT CCAT ATTAGGACATCTGC GT CCAT ATTAGGACATCTGC GT CCAT ATATGGGCAGCGAC TC CCAT ATATGGCCATGTAC	0.855 0.901 0.865 0.872 0.941 0.857 0.904 0.914 0.866 0.897 0.955 0.955 0.926 0.940 0.843 0.854 0.945 0.945 0.978 0.988

Additional information

17 out of 20 sequences are recognized by matrix family P$MADS.
20 out of 20 sequences are recognized by matrix family V$SRFF.

Save Matrix

In case you want to save the resulting matrix to your personal library, some more information has to be entered:

Matrix Identification
Matrix Identification	The matrix identification consists of a short name identifying the matrix (matrix name) and a more detailed description of the binding site (short description).
Family Information	Each matrix belongs to a so-called matrix family, where functionally similar matrices are grouped together in order to eliminate redundant matches by MatInspector. You can create a new family for your matrix by entering the name of the family or include your matrix into an existing user-defined matrix family by selecting a family from the list of available matrix families.
Extra Information	Here you can enter further information which will be stored in the References field of the matrix.

Matrix Identification

The matrix identification consists of

a short name identifying the matrix (matrix name)
and
a more detailed description of the binding site (short description).

Family Information

Each matrix belongs to a so-called matrix family, where functionally similar matrices are grouped together in order to eliminate redundant matches by MatInspector.

You can

create a new family for your matrix by entering the name of the family
or
include your matrix into an existing user-defined matrix family by selecting a family from the list of available matrix families.

Extra Information

Here you can enter further information which will be stored in the References field of the matrix.

References

MatDefine is described in:

Cartharius K, Frech K, Grote K, Klocke B, Haltmeier M, Klingenhoff A, Frisch M, Bayerlein M, Werner T (2005).
MatInspector and beyond: promoter analysis based on transcription factor binding sites.
Bioinformatics 21, 2933-2942.