stk, you can install it with pip:
$ pip install stk
Make sure you also install
rdkit, which is a dependency:
$ conda install -c conda-forge rdkit
stk is a Python library which allows construction and
manipulation of complex molecules, as well as automatic
molecular design, and the creation of molecular, and molecular property,
stk provides tools for constructing a variety of molecular
structures, including organic and metal-organic cages,
covalent organic frameworks, polymers and metal complexes, to name a
few. While additional molecular
structures are always being added,
stk provides tools which
allow users to easily specify new kinds of molecular structures,
no matter how simple or complex, in case something they want to build
is not built-in.
stk also makes it easy to specify which isomer
of a given structure should be built.
Here is a preview of some of the more complex structures
stk can construct
Automatic Molecular Design¶
stk provides an evolutionary algorithm, which can be used for
discovery of new molecules, which have properties desired by the user.
This evolutionary algorithm works with any molecule
construct, even those defined by users.
stk provides tools for the creation of molecular databases,
and storage of molecular property values.
stk comes with
support for the creation of MongoDB databases, which can be used to
store molecules constructed by users, or discovered by the
evolutionary algorithm. In addition, property values calculated for
those molecules can also be stored and retrieved from the database.
Molecular Database Visualization¶
stk has a sibling project called stk-vis, which is a
cross-platform application, which lets you connect to a database
stk, and view the molecules deposited into it.
stk-vis also shows you any molecular properties you deposited,
and lets you browse the building blocks of any constructed molecules.
It’s ideal for collaboration between multiple people, because one
person can deposit a molecule into the database, and another person
can immediately see and examine it. stk-vis is a stand-alone
application and does not require coding or
stk to use.
Features of stk-vis include
3D interactive molecular rendering.
Image of the 2D molecular graph.
Tabulation of any molecular properties deposited into the database.
Sorting molecules according to property values to quickly find ones with the best and worst properties.
You can list the building blocks of any constructed molecules. If those building blocks are also constructed molecules, you can list their building blocks as well, and so on.
Writing molecules to files.
You read more about stk-vis here:
A goal of
stk is to provide simple interfaces and require
minimal information from users to achieve basic and common tasks,
but also allow extensive customization and extension in
order to fulfill complex requirements.
Extensibility and Customization¶
Every part of
stk can be extended and customized in user
code, and every user-made extension is indistinguishable from
natively implemented features. This means users can use
stk to construct new classes of molecules, add new
kinds of molecular databases, and add or customize evolutionary
algorithm operations, all
without looking at
stk source code. All such extensions will
work with the rest of
stk as though they were part of the
stk is under active development. You can get alerted
when a new release comes out by going to the GitHub page and
click on the
watch button in the top right corner. Then select
Releases only from the dropdown menu.
Important features in the future will include:
Distributed Evolutionary Algorithms¶
Evolutionary algorithms are very simple to parallelize. It’s just a
matter of calculating each fitness value on a separate CPU core.
However, this idea can be taken further. Instead of calculating the
fitness function on a separate CPU core, calculate it on a separate
stk will support fitness functions, which instead
of calculating the fitness value locally, send the molecule over the
internet to a server, so that the server is responsible for calculating
the fitness value, which it then sends back to the client. You will
be able to keep adding servers as your computational requirements
increase, and the load will be distributed fairly between them.
This will allow continuous horizontal scaling of
More Molecular Structures¶
stk is always being expanded with new molecular structures.
If there is a specific kind of molecule you would like to be able
to construct with
stk, and you don’t feel like implementing
it yourself, simply create an issue on
https://github.com/lukasturcani/stk/issues, and describe what you would
like to build.
Something you might like to do first, is look at the
construction overview, which can give you a picture of how
stk goes about constructing molecules. It will also introduce
you to the basic concepts and types found within
stk. Next, the
basic examples, will allow
you to get a feel for how to use
stk. After that, examples of
molecular construction can be seen by looking at the documentation of
the different topology graphs. In general, you will find examples on
how to use a class, in that classes documentation.
Once you are comfortable with construction, you can start looking at
how to deposit and retrieve the molecules from databases. Finally,
stk provides multiple examples on how to use its evolutionary