McPromoter MM:II -- The Markov Chain Promoter Prediction Server
Massachusetts Institute of Technology

A statistical tool for the prediction of transcription start sites in eukaryotic DNA

NEW The Drosophila release 3 genome special issue of Genome Biology, including a genome-wide analysis of Drosophila core promoters and McPromoter predictions, has been published here! NEW

McPromoter is a program aiming at the exact localization of eukaryotic RNA polymerase II transcription start sites. We offer specific models designed for either human/vertebrate or D. melanogaster DNA. This service is free for non-commercial use and restricted to sequences up to 20 kb. The results will be mailed to you, so providing an email address is *mandatory*. We are always thankful for feedback -- especially if you obtain useful results...

Read the Q & A site first for information about the recent changes and instructions how to use McPromoter!
(Also contains links to the McPromoter papers and Uwe's thesis.)

Sensitivity and specificity are now at about 40-50% for both vertebrate and Drosophila at the transcript level .
Hint: If you don't get a prediction, have a look at the graphics that comes with the email to find optima that might have been missed by the pre-specified sensitivity level.

Note for Drosophila: If you want to retrieve predictions for a large fly region, use the GadFly analysis interface. Type in a chromosome arm under "Query sequence" (either 2L, 2R, 3L, 3R, 4, or X), and choose "promoter" as Program. This will give you all predictions in this arm with scores over 0.85, without any filtering for repeats or low-complexity. For specific regions, these predictions also show up in the FlyBase Genome Browser, if you choose "Dump view as TABLE" in the "display settings".

Submission by pasting a single sequence:

How to process your sequence ? forward only (default) forward & reverse

Which organism ? D. melanogaster human/vertebrate (default)

Sensitivity level ? Drosophila threshold Vertebrate threshold
highest (65%)
0.8 -.005
high (50%)
0.9 +.005
medium (35%)
0.95 +.015
low (20%) 0.98 +.025

Email address

Sequence name


... or Cut & Paste:
You should submit a single plain sequence using single letter nucleotides (A,C,G,T).
Warning: Long stretches of ambiguous symbols (> 50 bp) will be skipped; short stretches will be replaced randomly.

McPromoter was developed by Uwe Ohler at the University of Erlangen-Nuremberg, with help from Georg Stemmer, Andreas Kulicke. The software is copyright 1999--2001 by University of Erlangen-Nuremberg (Chair for Pattern Recognition). All rights are reserved.
Problems, Comments? email to Uwe Ohler
Neural networks courtesy of Stuttgart Neural Network Simulator SNNS

Our favourite user comment:
"I'll have a McPromoter, a pack of enhancers and a TATA box, to go!"

Web page & interface created by Moussa Sagna & Uwe Ohler.
Last update: 12.12.02