MEME System Release Notes

MEME version 3.0 -- December, 2000

• MEME enhancements
  • Hypertext output: MEME now reports its output in hypertext (HTML) format with appropriate internal hot links among the results and the self-contained documentation on how to interpret them. The alignments are color-coded by nucleic acid or amino acid category.
  • Direct MAST search: MEME motifs can now be used to search sequence databases using MAST simply by clicking on a button on the MEME HTML output document in an HTML-capable email reader or browser.
  • Direct BLOCKS search: MEME motifs can now be submitted to the BLOCKS multiple alignment processor by clicking on a button on the MEME HTML output document in an HTML-capable email reader or browser. This allows MEME motifs to be converted to LOGOS or trees, and to be used to search other databases of motifs.
  • New objective function: MEME searches for motifs that optimize the statistical significance of the log likelihood ratio of the occurrences of the motif.
  • E-values: MEME computes and reports the statistical significance of motifs as the E-value of the log likelihood ratio. This provides an objective measure of how likely the motif is to be biologically significant.
  • E-value stopping criterion: MEME will now stop when a motif whose E-value is above a user-given threshold is found. This guarantees that only motifs with a given E-value or better will be present in the output.
  • Handling reverse complement DNA strands: MEME now handles reverse complement DNA strands correctly with all model types: OOPS, ZOOPS and TCM.
  • Handling of ambiguous characters: MEME now handles ambiguous DNA and protein letters by converting them to the character "X". MEME treats the "X" character as "unknown", and correctly computes the probabilities of motif occurrences containing it.
  • Higher-order background models: MEME now allows the user to specify a Markov model of arbitrary order via a file of tuple frequencies. This appears to improve the ability to discriminate between biologically interesting DNA motifs and motifs that are artifacts of the higher order statistics of DNA sequence.
  • Multiple-alignment-based motif trimming: MEME defines a motif as a set of subsequences that can be correctly aligned without gaps. MEME can now trim the edges of motifs based on a local multiple alignment with gaps. MEME first determines the occurrences of the motif, then aligns each occurrence to the highest-scoring occurrence. These are combined into a multiple alignment, and MEME looks for a set of columns with no gaps. If a set of gapless columns at least wide is not found, MEME searches for a set with at most 1, 2, ... etc gaps until a set is found.
  • Prior distribution on the number of occurrences: MEME now places a prior on the number of occurrences that controls how strong the bias towards motifs with a given number of sites is. Using a prior improves the performance of MEME with DNA, where the megaprior heuristic (used by default with protein sequences), is not applicable.

• MAST enhancements
  • Combining DNA strands: MAST now combines the score of a site in a DNA sequence with the score of the corresponding site on the reverse complement strand. (The final score for the site is the maximum of the two scores.)
  • Scoring DNA strands separately: MAST still allows each strand of a DNA sequence to be treated as a separate sequence at the user's request.
  • Ignoring reverse complement DNA: MAST also allows the user to score only the given DNA strand, not scoring the reverse complement strand at all.
  • Background model: MAST now allows the user to specify the background residue frequencies used for computing E-values of scores.
  • Composition-adjusted statistics: MAST can now use a different random model for each target sequence, based on the letter composition of that sequence. This can greatly reduce erroneous matches due to biased sequence compostion.
  • New databases: The MAST website includes several new databases including "upstream" sequences for E. coli, B. subtilis and S. cerevesiae, and many complete genomes from GenBank.

• MEME and MAST enhancements:
  • Both MEME and MAST can be installed under the following operating systems:
    • Linux (various manufacturers)
    • SunSparc workstations (SunOS and Solaris operating systems)
    • Decalpha workstations (OSF/1 operating system)
    • Silicon Graphics workstations (IRIX version 5.3 operating system)
    • Intel Paragon parallel computers
    • Cray T3D parallel computers

MEME version 2.2 -- February 25, 1998