meme-explanation.html

来自「EM算法的改进」· HTML 代码 · 共 412 行 · 第 1/2 页

    <TR>
      <TD>Y</TD>
      <TD><FONT COLOR=TURQUOISE>TURQUOISE</FONT></TD> </TR>
  </TABLE>
</CENTER>
<p>
J. Kyte and R. Doolittle, 1982.
"A Simple Method for Displaying the Hydropathic Character of a Protein",
J. Mol Biol. 157, 105-132.

<p>
Summing the information content for each position in the motif gives
the total information content of the motif (shown in parentheses to the
left of the diagram).  The total information content is approximately
equal to the log likelihood ratio divided by the number of occurrences times
ln(2).
The total information content gives a measure of
the usefulness of the motif for database searches.
For a motif to be useful for database searches, it must as a rule contain at 
least log_2(N) bits of information 
where N is the number of sequences in the database being searched.  
For example, to effectively search a database containing 100,000 sequences
for occurrences of a <I>single</I> motif, the motif should have an IC of at
least 16.6 bits.  Motifs with lower information content are still useful when a 
family of sequences shares more than one motif since they can be combined
in <I>multiple</I> motif searches (using MAST).

<P>
<LI> <A NAME=consensus_doc2 HREF=#consensus1><H4>Multilevel 
  Consensus Sequence</H4></A>

The multilevel consensus sequence corresponding to the motif is an aid in 
remembering and understanding the motif.   It is calculated from the motif 
position-specific probability matrix as follows.  
Separately for each column of the motif, 
the letters in the alphabet are sorted in decreasing order by the probability 
with which they are expected to occur in that position of motif occurrences.  
The sorted letters are then printed vertically with the most probable letter 
on top.  Only letters with probabilities of 0.2 or higher at that position in 
the motif are printed.  As an example, the multilevel consensus sequence of 
motif 1 in the sample output is:
<PRE>
<B>Multilevel</B> <B>TTATGTGAACGACGTCACACT</B>
<B>consensus</B>  AA  T A G A GA     AA
<B>sequence</B>           T C TT     T
</PRE>
This multilevel consensus sequence says several things about the motif.
First, the most likely form of the motif
can be read from the top line as 
<TT>TTATGTGAACGACGTCACACT</TT>.
Second, that only letter <TT>A</TT> has probability more than 0.2 in
position 3 of the motif, both <TT>T</TT> and <TT>A</TT> have probability
greater than 0.2 in position 1, etc.
Third, a <I>rough approximation</I> of the motif can be made by converting the
multilevel consensus sequence into a 
<A HREF=#regular_expression_doc2>regular expression</A> for the motif.
<UL> 
<TT>[TA][TA]AT[GT][T][GA]A[AGT]C[GAC]A[CGT][GAT]TCACA[CAT][TA]</TT>
</UL>

<P>
<LI> <A NAME=sites_doc2 HREF=#sites1><H4>Occurrences of the Motif</H4></A>

MEME displays the occurrences (sites) of the motif in the training set.  
The sites are shown aligned with each other, and the ten sequence
positions preceding and following each site are also shown.
Each site is identified by the name of the sequence where it occurs,
the strand (if both strands of DNA sequences are being used), and the
position in the sequence where the site begins.  When the DNA strand
is specified, `+' means the sequence in the training set,
and `-' means the reverse complement of the training set sequence.
(For `-' strands, the `start' position is actually the position on the
<B>positive</B> strand where the site ends.)
The sites are listed in order of increasing statistical significance
(<I>p</I>-value).  The <I>p</I>-value of a site is computed from the the match 
score of the site with the <A HREF=#pssm>position specific scoring matrix</A> 
for the motif.  The <I>p</I>-value gives the probability of a random string
(generated from the background letter frequencies) having the same match
score or higher.  (This is referred to as the <B>position <I>p</I>-value</B>
by the MAST algorithm.)

<P>
<LI> <A NAME=diagrams_doc2 HREF=#diagrams1><H4>
Block Diagrams of Motif Occurrences</H4>
</A>
The occurrences of the motif in the training set sequences are 
shown with MAST-style block diagrams.  One diagram is printed for each
sequence showing all the occurrences of the motif in that sequence.
The sequences are sorted by the <B>lowest</B> <I>p</I>-value among all 
occurrences of the motif in a given sequence.
(The <I>p</I>-value of an occurrence is the probability of a single
random subsequence the length of the motif,
generated according to the 0-order background model, having a score
at least as high as the score of the occurrence.)
When the DNA strand is specified, `+' means the motif appears from left to
right on the sequence, and `-' means the motif appears from right to left 
on the complementary strand.
A sequence position scale is shown at the end of each table of block
diagrams.  Very long sequences are shown with thick lines connecting the
motifs and are <B>not</B> drawn to scale.

<P>
<LI> <A NAME=BLOCKS_doc2 HREF=#BLOCKS1><H4>
Motif in BLOCKS format or FASTA format</H4>
</A>

For use with 
<A HREF="http://blocks.fhcrc.org/blocks">BLOCKS tools</A>, 
MEME prints the occurrences of the motif in BLOCKS format. 
<P>
You can convert these blocks to 
PSSMs (position-specific scoring matrices), LOGOS (color representations
of the motifs), phylogeny trees and search them against a database of other
blocks by pasting everything from the "BL" line to the "//" line (inclusive)
into the  
<A HREF=http://blocks.fhcrc.org/blocks/process_blocks.html>
Multiple Alignment Processor.</A> 
If you include the <B>-print_fasta</B> switch on the command line, MEME prints
the motif sites in FASTA format instead of BLOCKS format.

<P>
<LI> <A NAME=pssm_doc2 HREF=#pssm1><H4>Position-Specific Scoring Matrix</H4></A>

The position-specific scoring matrix corresponding to the motif is printed
for use by database search programs such as MAST.  This matrix is a
log-odds matrix calculated
by taking 100 times the log (base 2) of the ratio <I>p/f</I> at each position in
the motif where <I>p</I> is the probability of a particular letter at that
position in the motif, and <I>f</I> is the background frequency of the
letter (given in the <A HREF=#command_doc>command line summary</A> section.)
This is the same matrix that is used above in computing the <I>p</I>-values
of the occurrences of the motif in the <A HREF=#sites_doc2>Occurrences of the
Motif</A> and <A HREF=#diagrams_doc2>Block Diagrams of Motif Occurrences</A>
sections.  
The scoring matrix is printed "sideways"--columns
correspond to the letters in the alphabet (in the same order as shown in
the simplified motif) and rows corresponding to the positions of the motif,
position one first.  The scoring matrix is preceded by a line starting with
"log-odds matrix:" and containing the length of the alphabet, width
of the motif, number of characters in the training set, the scoring 
threshold (obsolete) and the motif <I>E</I>-value.
<P>
<B>Note:</B> The probability <I>p</I> used to compute the PSSM
is <I>not</I> exactly the same as the corresponding value in the 
Position Specific Probability Matrix (PSPM).  
The values of <I>p</I> used to compute the PSSM take
into account the motif prior, whereas the values in the PSPM are just
the <I>observed</I> frequencies of letters in the motif sites.

<P>
<LI> <A NAME=pspm_doc2 HREF=#pspm1><H4>
Position-Specific Probability Matrix</H4></A>

The motif itself is a position-specific probability matrix giving,
for each position in the pattern, the observed frequency 
("probability") of each possible letter.  
The probability matrix is printed "sideways"--columns
correspond to the letters in the alphabet (in the same order as shown in
the simplified motif) and rows corresponding to the positions of the motif,
position one first.
The motif is preceded by a line starting with
"letter-probability matrix:" and containing the length of the alphabet, width
of the motif, number of occurrences of the motif, and the <I>E</I>-value of the
motif.
<p>
<b>Note:</b> Earlier versions
of MEME gave the posterior probabilities--the probability after applying
a prior on letter frequencies--rather than the observed frequencies.
These versions of MEME also gave the number of <I>possible</I>
positions for the motif rather than the actual number of occurrences.
The output from these earlier versions of MEME can be distinguished
by "n=" rather than "nsites=" in the line preceding the matrix.

<P>
<LI> <A NAME=regular_expression_doc2 HREF=#regular_expression1><H4>
Regular Expression</H4></A>
This is the <A HREF=#consensus_doc2>multilevel consensus</A> expressed as
a regular expression for convenience.  Regular expressions can
be used for searching for against sequences (using, for example,
<A HREF="http://nar.oxfordjournals.org/cgi/content/full/33/suppl_2/W262">PatMatch</A>)  
but the search accuracy will usually be better with the PSSM (using,
for example
<A HREF=http://meme.nbcr.net/meme/mast-intro.html>MAST</A>.)
MEME regular expressions are interpreted as follows: 
single letters match that letter; groups of letters in square brackets 
match any of the letters in the group.

<P>
<LI> <A NAME=motif-summary-doc2 HREF=#motif-summary><H4>
Motif Summary Tiling</H4></A>
The motif summary tiling is done using the same algorithm as used
by <A HREF=http://meme.nbcr.net/meme/mast-intro.html>MAST</A>.
The motif occurrences shown in the motif summary
<B>may not be exactly the same as those reported in each motif section</B>
because only motifs with a position <em>p</em>-value of 0.0001 that
don't overlap other, more significant motif occurrences are shown.
The format of the machine readable motif-summary is:
<pre>
[sequence_name combined_<em>p</em>-value number_of_motif_occurrences [motif_number start_of_motif position_<em>p</em>-value]+]+
</pre>
See the documentation for 
<A HREF=http://meme.nbcr.net/meme/mast-output.html>MAST output</A> for the definition of position and
combined <em>p</em>-values.

</UL>