Hi RDKit'ers,
I've just released new versions of chemfp. Version 1.6 is the no-cost/open
source version, and 3.4 is the commercial version.
The goal of chemfp 1.6 is to provide a good performance baseline for evaluating
new Tanimoto search programs. This release is about 10-20% faster than chemfp
1.5. I believe it is the fastest no-cost/open source search program available
for CPUs. It can be installed with:
python -m pip install chemfp --upgrade
However, it only supports Python 2.7, which means it's only useful for legacy
RDKit installations.
On the other hand, chemfp 3.4 supports Python 2.7 and Python 3.6+, is faster,
and has many more features than chemfp 1.6, including a "toolkit" I/O
abstraction to handle reading/writing structures to/from a file or a string.
The comprehensive documentation is at:
https://chemfp.readthedocs.io/en/latest/
See the "What's New" section for a list of changes over time.
Chemfp 3.4 is available with academic and commercial (proprietary) licenses,
both for pre-compiled binary packages which work on most Linux-based OSes, and
for an option also get source code access. Open source licensing is still
available, as the most expensive option.
With this release I am making those pre-compiled packages available for anyone
to download under a base license agreement for in-house use, which limits or
restricts access to some features:
• fingerprint arenas may not be larger than 50,000 fingerprints;
• in-memory arena searches may not have more than 50,000 queries or
targets;
• FPS searches may not have more than 20 queries;
• Tversky search is disabled;
• writing FPB files is disabled.
These can be enabled with a valid license key.
There's a special waiver which lets you use this version of chemfp to generate
FPS files for any purpose, including commercial use. (This also gives a way for
chemfp 1.6 to be a useful benchmark baseline even with new data sets and
fingerprint types.)
See https://chemfp.com/license/ for details, or to get started:
python -m pip install chemfp -i https://chemp.com/packages/ --upgrade
A one-year no-cost license key for the pre-compiled packages is also now
available for academics.
Contact me if you have questions or are interested in evaluating or licensing
chemfp.
Andrew
da...@dalkescientific.com
___
Rdkit-discuss mailing list
Rdkit-discuss@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
