.. _highlights_2019_Apr:

******************************
OEToolkits 2019.Apr
******************************

FastROCS TK: Color calculation on GPU
================================================================================

Changes to the **FastROCS** algorithm, including porting the static color
calculation to GPU, have resulted in a more than 2x performance improvement of **FastROCS** on some GPU systems.
Benchmarking was carried out on AWS GPU instances and compared to the 2018.Oct
version of **FastROCS TK**:

.. figure:: images/2019-Apr/2018-2019_comparison_aws.png
   :width: 400pt
   :align: center

   **Performance comparison between the 2018.Oct release and the 2019.Apr release using
   OpenEye's standard benchmark dataset of 14M conformers on AWS.**

For some systems, the standard benchmark dataset of 14M conformers is not large
enough to fully saturate the GPUs. To demonstrate the potential
performance of larger databases, a 2 billion conformer subset of
Enamine was searched on suitable AWS instances. FastROCS searched the
2 billion conformer database on an AWS P3 instance with 8x Nvidia
Tesla V100 GPUs at a rate of over 44 million overlays per second:

.. figure:: images/2019-Apr/AWS-Enamine-2bn.png
   :width: 300pt
   :align: center

   **Performance of a 2 billion conformer subset of Enamine on suitable AWS instances.**

OEChem TK: OEZ compressed file format
================================================================================

A new compressed file format with the extension ``oez`` has been added to
**OEChem TK**.
This format uses the `Zstandard <https://facebook.github.io/zstd/>`_ algorithm
to efficiently compress individual multi-conformer molecules, making it well-suited
to work with :ref:`OEMolDatabase <OEChem_OEMolDatabase>` for fast read-only random access to molecules.
See more details in the :ref:`section_molreadwrite_compressed` section of the
**OEChem TK documentation**.

OEGrapheme TK: Ramachandran plot
================================================================================

Following our tradition of providing easily comprehensible visualization of
complex data, **OEGrapheme TK** can now depict Ramachandran plots ([Ramachandran-1963]_).
These plots provide an easy way to examine the distribution of backbone dihedral angles in a particular protein.
For instance, structure quality can be assessed by identifying residues whose
backbone :math:`\psi` are outside energetically accessible regions.
**Grapheme TK** uses the data and classifications available in **OEBio TK**,
which were extracted from the open-source
Computational Crystallography Toolbox [Grosse-Kunstleve-2002]_.

.. only:: oehtml

   .. raw:: html

      <p align="center">
      <b > Hovering over the red circles (on the top right plot) reveals the name of the outlier residues</b>
      <p>

      <p align="center">
      <object data="../_images/ramagrid2img-2rox.svg" type="image/svg+xml"  width = "800" ></object>
      <p>

      <p align="center">
      <b>Examples of visualizing the Ramachandran plots of various residue types in Grapheme TK (PDB: 2ROX)</b>
      <p>

  .. this is necessary to copy images into build subdirectory

  .. image:: images/2019-Apr/ramagrid2img-2rox.svg
     :width: 1pt

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

* **OpenEye** currently uses `OpenJDK8 <https://openjdk.java.net/>`_ to
  build its Java packages. We are currently exploring support for OpenJDK 11.

* The 2019.Apr release skips Java support for macOS. For Java support on RHEL6,
  please contact support@eyesopen.com.

* This release adds support for macOS 10.14. OS X 10.11 is no longer supported.

* This release adds support for Ubuntu 18. Ubuntu 14 is no longer supported.

* Visual Studio 2017 support has been added for C++ and C#.
  This is the last release to support Visual Studio 2013,
  as well as 32-bit Windows for C++ and C#.

* Full support has been added for Python 3.7.
  Python 3.5 for Windows is no longer supported.

* The following prerequisites have been introduced
  so that **OpenEye** can implement algorithms in a more modern and
  efficient C++11 style:

  * GCC 4.8 is now the oldest targeted compiler for toolkits.
    RHEL6 with the default GCC 4.4 compiler is no longer supported.

  * The minimum recommended GNU `binutils <https://www.gnu.org/software/binutils/>`_
    version is 2.26 for the C++ single-build Linux packages.

  * The minimum CMake version required to build the C++ examples is now 3.1.3
    on all platforms.
