Clustering Steps
The basic steps are as follows:
- (IF data is in MOL format:) Convert the MOLfile to Daylight TDT format
with "mol2tdt". This is a contributed program at
$DY_ROOT/contrib/src/applics/convert/molfiles/mol2tdt.c
You have to compile the program with:
cd DY_ROOT/contrib/src
make
cd DY_ROOT/contrib/src/applics/convert/molfiles
make mol2tdt
There may be a newer version of the program at our ftp
site daylight.daylight.com, directory
/var/spool/ftp/pub/contrib/src, filename "mol2smi.c".
- Fingerprint (fingerprint program) the TDT file.
- Compute nearest neighbors (nearneighbors).
- Use jpscan to predict clustering results and select clustering
parameters.
- Cluster with chosen parameters (jarpat).
- Add variances to clusters (listclusters -v).
- (Optionally) display clustering results with showclusters.
- Create a Daylight database (datatypes database and regular
database). (Use thormake or sthorman.) See
Creating a Daylight Database for guidance.
- Load database with datafiles. Use output of fingerprint
program plus output of listclusters program (2 loads, merging).
(thorload or sthorman).
- Load pool (merlinload or sthorman) to view with Merlin.
This is a brief description of the process. Please refer to the
Daylight Administration manual for more guidance, and the Daylight
Theory manual (Thor/Merlin especially), and the Clustering manual.
