6 Oct 1998 KEvin Karplus
    The "abstracts" have been moved to /projects/compbio/papers/casp3/abstracts

Sun Jun 21 17:38:03 PDT 1998	Melissa Cline

This directory serves as a repository for CASP3 predictions to be
submitted, acknowledgements of those predictions, and programs for
assembling predictions.  Note that all that should be here is
predictions that will be submitted.  Interesting looking things that we
might want to submit should stay in the target directory until we decide
to submit them.  This is to keep the submit direcotries from getting so
cluttered that we can't remember what we did and didn't submit.

Predictions may be submitted via email to
submit@predictioncenter.llnl.gov or via the web at
http://PredictionCenter.llnl.gov/casp3/submit/casp3-ver.html

Here are some important files found in this directory:
- registration.fold: CASP3 registration ID for the fold recognition team.
- methods.fold: general description of target98 fold recognition.
- remark: sample remark file with the names of the authors.
- AlForNewFold: alignment record to be used when predicting a new fold.
- fasta2al: program for producing an AL record from a FASTA-format alignment.
- al2fasta: program for converting a prediction back into a FASTA alignemnt.
- format_prediction: script for assembling a prediction record out of
  its component files..

Here are the steps for assembling a single prediction:

1. make a directory submit/<target>.

2. In the directory submit/<target>, write a file results.<target>
   summarizing the results on that target: scores, best hits, quality
   of best alignments.  Optionally, write a file methods.<target>
   describing any special methods used on this prediction. 

3. Create a methods file as follows:
   cat ../methods.fold results.<target> > methods.<target>

4. If not predicting a new fold, copy the prediction alignment to
   the directory of step 1.

   Create an AL record from the alignment as follows:
   ../fasta2al <-candidate align_filename> <-template templateName> \
	       [-target targetName] [-workingcandidate align_2_filename] \
	       > <align_name.al>
   All switches can be abbreviated to two characters (eg: -ca instead
   of -candidate).  
   If the alignment is pairwise, the target name will be deduced from the 
   alignment.  Otherwise, the -target switch is required to specify which 
   sequence to include in the pairwise alignment with the template sequence.

   fasta2al will compare the target sequence in the alignment to the
   sequence in the file casp3/<target>/<target>.seq.  If the two
   versions of the sequence are different, it will print out an error 
   message and halt.  fasta2al will also compare the template sequence
   from the alignment with the SEQRES sequence produced by the pdb
   summary tool  If the sequence versions are different, it will print
   out a warning message and use SAM to reconcile differences in the
   sequences.  Any missing residues will be added to the target/template
   alignment as an insert in the template sequence, a '.' in the target
   sequence.  If the -workingcandidate (abbrev: -wo) switch is
   specified, this alignment will be printed to the specified filename. 

   sample command line: ../fasta2al -cand 1prcC-t54-global.pw.a2m -te 1prcC \
   				> 1prcC-t54.global.pw.al

5. Assemble a remark file with whatever contents you wish.  Remark
   records generally don't affect the assessment of the prediction, and
   aren't to be distributed at the meeting with the methods records.
   Good things for the remarks section are a prettyalign of the
   alignment submitted, group name, name of authors, and any previous
   predictions that this prediction supercedes.

6. Assemble a prediction file from the components with format_prediction
   as follows:

   ../format_prediction <-target targetnum> <-methods methods_file>  \
			<-al al_file> [-remark remark_file] > <submit_file>

   sample command line: ../format_prediction -target T0054 -methods \
			methods.t54 -remark remark.t54 \
			-al 1prcC-t54-global.pw.al >submit.t54

   If making a new fold prediction, use <-al ../ALForNewFold>.

   Note that the targetnumber should probably be as it appears in the
   CASP3 documentation: with a capital T and a bunch of zeros, eg. T0061
   rather than t61.

7. This might not be necessary, but is safer; if making a new fold
   prediction, edit the submit file and change the first line from
   "PFRMAT AL" to "PFRMAT TS".

8. If not making a new fold prediction, verify the alignment with
   al2fasta as follows:
   
   ../al2fasta -output <align_file.a2m> [-template template_name] \
	       < <submission_file>

   sample command line: ../al2fasta -output t54-1prcC.a2m < submit.t54

   The -template switch is provided because at some later time, we will
   have the capability of concatenating multiple predictions.  This 
   feature is provided for extracting the alignment of one particular
   template structure.

   Do a prettyalign of align_file.a2m and the original alignment to
   verify that the alignments are consistent - that the beginning and
   end of the match columns are the same.

9. Mail the submit file to the email address listed above.  Shortly 
   afterwards, you will be mailed an acknowledgement.  Save the
   acknowledgement in the submit directory.

Extra details:

domain hits off different templates
-----------------------------------
When making non-overlapping predictions to different targets (eg. domain
hits), the AL records should be concatenated.  The overall model should
start with the first domain hit and include a normal AL record.  After
the TER record, the next domain hit should begin with the PARENT record
and continue down through the TER record.  This is achieved by
concatenating AL files in the correct order.

chains in the target sequence
-----------------------------
When the target sequence is broken into one or more chains, the chain ID
must appear in the AL record immediately before the target sequence
number.  In other words, to specify an alignment to target chain A, instead of
A 4 I 71
there should be
A A4 I 71

predictions detailing protein interactions
------------------------------------------
See t66 for an example.  A TS record should be submitted in place of an
AL record.  There's a conversion tool available via the CASP3 web pages.
The coordinates of both proteins should be delimited with a single pair
of PARENT and TER records.  All parents involved in the interaction
should be listed on this single PARENT line.

Multiple Models
---------------
To submit multiple models, a separate submission of each model 
is necessary.  You must also alter the MODEL {1,2,3,4,5} field
accordingly.