Transcription Unit Predictions for E. coli K-12

The file "TU.txt" contains our transcription unit (TU) predictions for E. coli
K-12 using the methods described in "Predicting bacterial transcription units 
using sequence and expression data" (Bockhorst et at. ISMB 2003).  This file 
contains a set of TU records separated blank lines.  Each TU record consists of
seven lines.  These lines identify the name of the TU, the location and strand 
of the TU, the genes contained in the TU and the evidence used to predict the 
TU boundaries.

The predictions of our TU model were constrained so that all of the regulatory
elements in our training set appear in a TU (note that this contrasts with 
the methodology used in the article).  For example, each of the known 
operons in our training set is associated with a TU prediction.  A result of
this is that certain cases, such as overlapping operons, are reported here
that would be infeasible for the TU model as described in the article to 
predict.  

Here is an example TU record:

TU.5.0
12047	15346	FORWARD
dnaK dnaJ 
first-gene-source:        [PROMOTER] [START-OF-RUN] [START-OF-OPERON]
TSS-source:               [PROMOTER]
last-gene-source:         [END-OF-OPERON]
TES-source:               [PREDICTION]

The first line contains the name of the TU as "TU.x.y" where x identifies the
run the TU is in and y indexes TUs within a run.

The second line indicates the location of the TU as
<5' position> <3' position> <strand>
<5' position> indicates the position of the first base in the TU, <3' position>
indicates the position of the last base in the TU and <strand> indicates 
whether the TU lies on the forward (FORWARD) or reverse (REVERSE) strand.

The third line gives that name of all genes wholly contained within this
TU that are on the same strand

The fourth line indicates the evidence used for predicting the first gene 
of the TU. There are six different cases that cause a gene to be the
first gene of a predicted TU. 
(i)   [START-OF-OPERON] The gene is the first gene of a known operon.
(ii)  [PROMOTER] A promoter lies in the non-coding region (NCR) upstream of 
      the gene.
(iii) [START-OF-RUN] The gene is the first gene of a run.
(iv)  [UPSTREAM-OPERON] The upstream neighbor gene is the last gene in a known
      operon.
(v)   [UPSTREAM-TERMINATOR] A terminator lies in the NCR following the
      upstream neighbor.
(vi)  [PREDICTION] TU model predcits gene to be the first gene in TU based on
      DNA sequence and expression data.
One of these cases applies to every TU record and in general multiple cases
can apply to a single record however, [PREDICTION] only applies if none of the
other cases apply.

The fifth line indicates the evidence associated with the TSS.  There are
three possible cases and exactly one applies to every TU:
(i)   [PROMOTER] TSS is based on known promoter
(ii)  [PREDICTION] TSS is based on prediction of TU model
(iii) [GENE-START] TSS refers to the first base of the first gene in the TU.
The last case is used when the first gene of the TU is the first gene of
a TU (based on case (i), (iv), or (v) of line three) but that the space
of the non-coding region is too small for our TU model to predict a TSS.

The sixth line indicates the evidence used for predicting the last gene 
of the TU. As for the first gene, there are six different cases that cause a 
gene to be the last gene of a predicted TU. 
(i)   [END-OF-OPERON] The gene is the last gene of a known operon
(ii)  [TERMINATOR] A terminator lies in the NCR downstream of the gene
(iii) [END-OF-RUN] The gene is the last gene in a run
(iv)  [DOWNSTREAM-OPERON] The nearest downstream neighbor is the first gene of
       a known operon
(v)   [DOWNSTREAM-PROMOTER] A promoter lies in the NCR upstream of the
      downstream neighbor gene.
(vi)  [PREDICTION] TU model predicts gene to be the last gene in a TU based on
       DNA sequence and expression data.

The seventh line indicates the evidence associated with the TES.  As with the
TSS, there are three possible cases and exactly one applies to every TU:
(i)   [TERMINATOR] TES is based on known terminator
(ii)  [PREDICTION] TES is based on prediction of TU model
(iii) [GENE-END] TES refers to the last base of the last gene in the TU.
The last case is analogous to case (iii) for the TSS and comes up when
evidence suggests the gene to be the last gene in a TU but the downstream NCR 
is too small for our TU model to make a prediction.