.. _toolkit_release_highlights_2017_Jun:

******************************
OEToolkits 2017.Jun
******************************

Shape TK: Hermite
================================================================================

The 2017.Jun release introduces the :ref:`OEHermite <OEShape_OEHermite>` class, which
allows a molecule to be expanded into a Hermite representation.  An initial
suggested area of application is the representation of protein shape. The
advantages of representing proteins by Hermites include:

* smooth representations that capture any level of detail, from the
  atomistic (many coefficients, see the picture on the right) to the very
  coarse (few coefficients, as in the middle picture),

* the ability to store these representations in a compact manner,

* the ability to transform these representations easily, and 

* very fast overlap calculations between proteins. 

Below are examples of low- and high-resolution Hermite representation
images of dihydrofolate reductase (DHFR):

.. figure:: images/2017-Jun/hermite-figure.*
   :width: 650pt
   :align: center
   :alt: Hard and soft shape gaussians
  
   **Exact shape versus Hermite Representation**

**OpenEye** will explore the potential of Hermite representations; we invite
users to do likewise with this exciting new addition to the OpenEye toolkits.
See the :ref:`shapetk_hermite_section` chapter of the **Shape TK**
documentation for more information.

OEDepict TK and Grapheme TK: Interactive SVG images
================================================================================

Scalable Vector Graphics (SVG) image format is dominant in web applications due
to its wide browser support and scalability.  An SVG image contains a list of
graphical objects that can be grouped, styled, and transformed together.
This enables the generation of interactive images defined in the SVG XML
elements or via scripting that accesses the SVG Document Object Model (DOM).

The new :term:`Preliminary API` available in **OEDepict TK** provides low-level
functionality that enables grouping of graphical elements in SVG images that
can be manipulated externally by web developers.
High-level API entrypoints are also available to generate self-contained SVG
images that show or hide different elements of the image based on mouse events.

.. only:: oehtml

   .. list-table:: **Example of interactive SVG images (PDB: 4HWR)**
      :widths: 400 400

      * - Hover the mouse over any ligand atom to display its b-factor
        - Click on any residue circle to display ligand atoms it interacts with

      * - .. raw:: html

             <p align="center">
             <object data="../_images/bfactor-4HWR.svg" type="image/svg+xml"  width = "400" ></object>
             <p>

        - .. raw:: html

             <p align="center">
             <object data="../_images/RenderContactMap-4HWR.svg" type="image/svg+xml"  width = "400" ></object>
             <p>

   .. these are necessary to copy images into build subdirectory

   .. image:: images/2017-Jun/bfactor-4HWR.svg
      :width: 1pt

   .. image:: images/2017-Jun/RenderContactMap-4HWR.svg
      :width: 1pt


For more information see the :ref:`chapter_svg` chapter of the **OEDepict TK**
documentation.

GraphSim TK: Fast Fingerprint Similarity Search
================================================================================

**OEGraphSim TK** now provides improved fingerprint search capability,
allowing users to perform molecule similarity searches on large databases
up to 10x faster than previous versions.

Fingerprint methods are used for the computation of similarity scores.
When searching large virtual databases, performance can be problematic.
To address this issue, **OEGraphSim TK** now uses :term:`popcount`, a
hardware instruction available on modern CPUs, to provide significant
performance improvements for computing similarity.

The new fingerprint search can be performed in two modes: *in-memory* and
*memory-mapped*.  The in-memory mode involves preloading all fingerprints
into memory and performing the search.  While this represents the
fastest way to perform similarity searches once the fingerprints are loaded,
searches are limited by memory availability.  The memory-mapped mode has no
load time penalty or memory limitation but the search itself takes more time.

.. list-table::
   :widths: 400 400

   * - fingerprint generation performance
     - fingerprint search performance


   * - .. image:: images/2017-Jun/PerformanceMakeFastFP.*
          :width: 300 pt

       * *generating 1M 4096-bit long tree fingerprints*
       * *benchmarked on m4.10xlarge AWS instance*
       * *multi-threaded fingerprint generation*

     - .. image:: images/2017-Jun/PerformanceFastFP.*
          :width: 300 pt

       * *128M 4096-bit long tree fingerprints = 64GB data*
       * *benchmarked on m4.10xlarge AWS instance*
       * *single-threaded search on CPU*

See the :ref:`oegraphsim_chapter_examples` chapter of the **OEGraphSim TK**
documentation for more information and examples using the new
:ref:`OEFastFPDatabase <OEGraphSim_OEFastFPDatabase>` class.

General Notices
================================================================================

* On Linux, Python ``single-build`` distributions are now installed by default
  when using the meta-installer via ``pip install openeye-toolkits``.
  If a problem is encountered with the ``single-build``
  packages, users can set the environment variable ``OE_PIP_ARCH=old``
  before the ``pip install`` step.

* Python 3.6 is supported in the 2017.Jun release.

* This is the last release to support 32-bit Java.

* This is the last release to support Red Hat Enterprise Linux version 5.
  As of March 31, 2017, RHEL5 has reached end-of-production support.

.. note::

   **OpenEye** continues to monitor and collect feedback from our customers
   on the need to support Python 2 wrappers.  Our current plan is to
   phase out support for Python 2 on all platforms by the 2017.Oct release.
   As this is a substantial change for our customers, we are willing
   to help with migration.  Please contact support@eyesopen.com for
   more details.
