STREME outputs a tab-separated values (TSV) file ('sequences.tsv') containing one line for
each sequence with a site whose score passes the motif's match threshold for each motif discovered by STREME.
The lines are grouped by motif, and groups are separated by a line
starting with the character "#".
The first line in the file contains the (tab-separated) names of the fields.
The names and meanings of each of the fields are described in the table below.
field
name
contents
1
motif_ID
The name of the motif uses the IUPAC codes for nucleotides or proteins.
Letters representing multiple nucleotides are used in nucleotide motif
positions where several nucleotides are favored.
The name of the motif is <index>-<consensus>,
where <index> is the rank of the motif according to P-value or Score,
and <consensus> is an approximation of the motif by an IUPAC sequence.
2
motif_ALT_ID
The alternate name of the motif is STREME-<index>,
where <index> is the rank of the motif according to P-value or Score.
3
motif_P-value
The p-value of the motif based on applying the appropriate statistical test
to the test set sequences. It is not adjusted for the number of motifs reported by STREME.
If STREME reports a single motif, then
the p-value is an accurate estimate of
the statistical significance of the motif as long as the length
distributions of the positive and negative sequences are essentially the same.
However, if STREME reports more than one motif, the p-value does NOT
completely account for multiple testing, and you should use the E-value
for assessing whether a motif is truly statistically significant.
motif_Score
The Score is the unadjusted p-value of the motif based on the appropriate test
applied to the training set sequences.
Since the Score is not adjusted for multiple tests, it cannot
be used to determine the statistical significance of the motif.
4
seq_ID
The ID of the sequence.
5
seq_Score
The seq_Score of a sequence is its maximum motif match score over all sequence positions. The motif match score of a position in a sequence is
computed by summing the appropriate entry from each column of the
position-dependent scoring matrix that represents the motif.
5
seq_Class
Whether the sequence is a true positive, 'tp', or a false positive, 'fp'.
6
is_holdout?
Whether the sequence was in the holdout set, '1', or not, '0'.
The name of the motif uses the IUPAC codes for nucleotides or proteins.
Letters representing multiple nucleotides are used in nucleotide motif
positions where several nucleotides are favored.
The name of the motif is <index>-<consensus>,
where <index> is the rank of the motif according to P-value or Score,
and <consensus> is an approximation of the motif by an IUPAC sequence.
The sequence logo of the motif.
The rules for construction logos are given in
the Description section of the documentation for the MEME Suite utility
ceqlogo.
The sequence logo of the reverse complement motif.
The rules for construction logos are given in
the Description section of the documentation for the MEME Suite utility
ceqlogo.
Click on the blue symbol below to reveal options allowing you
to submit this motif to another MEME Suite motif analysis program, to download this
motif in various text formats, or to download a sequence "logo" of
this motif PNG or EPS format.
Supported Programs
Tomtom
Tomtom is a tool for searching for similar known motifs.
[manual]
MAST
MAST is a tool for searching biological sequence databases for
sequences that contain one or more of a group of known motifs.
[manual]
FIMO
FIMO is a tool for searching biological sequence databases for
sequences that contain one or more known motifs.
[manual]
GOMo
GOMo is a tool for identifying possible roles (Gene Ontology
terms) for DNA binding motifs.
[manual]
SpaMo
SpaMo is a tool for inferring possible transcription factor
complexes by finding motifs with enriched spacings.
[manual]
This plot shows the positional distribution of the best match to the motif in the positive training sequences.
Only matches with scores at least the score threshold are considered.
The plot is smoothed with a triangular function whose width is 5% of the maximum positive training sequence length.
The position of the dotted vertical line indicates whether the sequences were
aligned on their left ends, centers, or right ends, respectively.
This histogram shows the distribution of the number of matches to the motif in the positive training sequences with at least one match.
Only matches with scores at least the score threshold are considered.
The Score is the unadjusted p-value of the motif based on the appropriate test
applied to the training set sequences.
Since the Score is not adjusted for multiple tests, it cannot
be used to determine the statistical significance of the motif.
For determining if a motif is statistically significant, you should use the
value in the E-value column. If there is no E-value column, that
means that either the positive or negative hold-out set would have been too small
(fewer than 5 sequences).
For very small sequence sets, it is not practical for STREME to compute an accurate E-value.
In that case, you can determine if your motif is significant by running STREME twenty or more
times on shuffled versions of your positive dataset,
and seeing if the Score is always larger than the Score using the original sequences.
You can make shuffled sequence datasets using the MEME Suite command-line utility
fasta-shuffle-letters) if you
have installed the MEME Suite on your own computer.
The statistical test used in computing the Score is either the Fisher Exact Test,
the Binomial Test, or the Cumulative Bates distribution. (See Inputs and Settings
for the particular test being used.) The Fisher Exact Test and the Binomial Test both
estimate the enrichment of the motif in the positive sequences compared to the
the negative sequences.
(The Binomial Test is used when the positive and negative sequences have different average lengths.)
The Cumulative Bates distribution measures the tendency
of motif to be near the center of the input sequences.
The p-value of the motif based on applying the appropriate statistical test
to the test set sequences. It is not adjusted for the number of motifs reported by STREME.
If STREME reports a single motif, then
the p-value is an accurate estimate of
the statistical significance of the motif as long as the length
distributions of the positive and negative sequences are essentially the same.
However, if STREME reports more than one motif, the p-value does NOT
completely account for multiple testing, and you should use the E-value
for assessing whether a motif is truly statistically significant.
The statistical test used in computing the p-value is either the Fisher Exact Test,
the Binomial Test, or the Cumulative Bates distribution. (See Inputs and Settings
at the bottom of this document for the particular test being used.)
The Fisher Exact Test and the Binomial Test both
measure the enrichment of the motif in the positive test sequences compared to the
the negative test sequences.
(The Binomial Test is used when the positive and negative sequences have different average lengths.)
The Cumulative Bates distribution measures the tendency
of motif to be near the center of the sequences.
The E-value is an accurate estimate of
the statistical significance of the motif as long as the length
distributions of the positive and negative sequences are essentially the same.
The E-value is the p-value multiplied by the number of motifs reported by STREME.
It is an estimate of the number of motifs that would be found with enrichment
as high as this motif in shuffled versions of your positive sequences.
The score threshold for determining if a potential site is a match
to the motif. The same threshold is applied when determining matches in
the training and test sequences. The threshold is in bits.
The match score of a position in a sequence is determined by converting
the motif to a base-2 log-odds matrix using the formula log2(prob[a][i]/background[a]).
Here, prob[a][i] is the probability of the letter 'a' at position 'i' of the motif,
and background[a] is the probability of the letter 'a' according to the background.
The names of the files containing the positive (primary)
and negative (control) sequences input to STREME.
If you did not provide a file containing the negative (e.g., control)
sequences, STREME created them using N-order shuffling.
0-order shuffling preserves 1-mer frequencies (i.e., the letter frequencies),
1-order shuffling preserves 2-mer frequencies, etc.
If you use STREME in your research, please cite the following paper:
Timothy L. Bailey,
"STREME: accurate and versatile sequence motif discovery",
Bioinformatics, Mar. 24, 2021.
[full text]
