The framework uses Google TensorFlow, along with scikit-learn, for expressing neural networks for deep learning.It also makes use of the RDKit python framework, for performing more basic operations on molecular data, such as converting SMILES strings into molecular graphs.The framework is now in the alpha stage, at version 0.1.As the framework develops, it will move toward implementing more models in TensorFlow, which use GPUs for training and inference.This new open source framework is poised to become an accelerating factor for innovation in drug discovery across industry and academia.

Another unique aspect of DeepChem is that it has incorporated a large amount of publicly-available chemical assay datasets, which are described in Table 1.

DeepChem Assay Datasets

Table 1:The current v0.1 DeepChem Framework includes the data sets in this table, along others which will be added to future versions.

Metrics

The squared Pearson Correleation Coefficient is used to quantify the quality of performance of a model trained on any of these regression datasets.Models trained on classification datasets have their predictive quality measured by the area under curve (AUC) for receiver operator characteristic (ROC) curves (AUC-ROC).Some datasets have more than one task, in which case the mean over all tasks is reported by the framework.

Data Splitting

DeepChem uses a number of methods for randomizing or reordering datasets so that models can be trained on sets which are more thoroughly randomized, in both the training and validation sets, for example.These methods are summarized in Table 2.

DeepChem Dataset Splitting Methods

Table 2:Various methods are available for splitting the dataset in order to avoid sampling bias.

Featurizations

DeepChem offers a number of featurization methods, summarized in Table 3.SMILES strings are unique representations of molecules, and can themselves can be used as a molecular feature.The use of SMILES strings has been explored in recent work.SMILES featurization will likely become a part of future versions of DeepChem.

Most machine learning methods, however, require more feature information than can be extracted from a SMILES string alone.

DeepChem Featurizers

Table 3:Various methods are available for splitting the dataset in order to avoid sampling bias.

Supported Models

Supported Models as of v0.1

Table 4: Model types supported by DeepChem 0.1

A Glimpse into the Tox21 Dataset and Deep Learning

The Toxicology in the 21st Century (Tox21) research initiative led to the creation of a public dataset which includes measurements of activation of stress response and nuclear receptor response pathways by 8,014 distinct molecules.Twelve response pathways were observed in total, with each having some association with toxicity.Table 5 summarizes the pathways investigated in the study.

Tox21 Assay Descriptions

Table 5:Biological pathway responses investigated in the Tox21 Machine Learning Challenge.

We used the Tox21 dataset to make predictions on molecular toxicity in DeepChem using the variations shown in Table 6.

Model Construction Parameter Variations Used

Table 6:Model construction parameter variations used in generating our predictions, as shown in Figure 1.

A .csv file containing SMILES strings for 8,014 molecules was used to first featurize each molecule by using either ECFP or molecular graph convolution.IUPAC names for each molecule were queried from NIH Cactus, and toxicity predictions were made, using a trained model, on a set of nine molecules randomly selected from the total tox21 data set.Nine results showing molecular structure (rendered by RDKit), IUPAC names, and predicted toxicity scores, across all 12 biochemical response pathways, described in Table 5, are shown in Figure 1.

Expect more from DeepChem in the Future

The DeepChem framework is undergoing rapid development, and is currently at the 0.1 release version.New models and features will be added, along with more data sets in future.You can download the DeepChem framework from github.There is also a website for framework documentation at deepchem.io.

Microway offers DeepChem pre-installed on our line of WhisperStation products for Deep Learning. Researchers interested in exploring deep learning applications with chemistry and drug discovery can browse our line of WhisperStation products.

References

1.) Subramanian, Govindan, et al. “Computational Modeling of β-secretase 1 (BACE-1) Inhibitors using Ligand Based Approaches.” Journal of Chemical Information and Modeling 56.10 (2016): 1936-1949.
2.) Altae-Tran, Han, et al. “Low Data Drug Discovery with One-shot Learning.” arXiv preprint arXiv:1611.03199 (2016).
3.) Wu, Zhenqin, et al. “MoleculeNet: A Benchmark for Molecular Machine Learning.” arXiv preprint arXiv:1703.00564 (2017).
4.) Gomes, Joseph, et al. “Atomic Convolutional Networks for Predicting Protein-Ligand Binding Affinity.” arXiv preprint arXiv:1703.10603 (2017).
5.) Gómez-Bombarelli, Rafael, et al. “Automatic chemical design using a data-driven continuous representation of molecules.” arXiv preprint arXiv:1610.02415 (2016).
6.) Mayr, Andreas, et al. “DeepTox: toxicity prediction using deep learning.” Frontiers in Environmental Science 3 (2016): 80.

The post DeepChem – a Deep Learning Framework for Drug Discovery appeared first on Microway.

This short tutorial explains the usage of the GPU-accelerated HOOMD-blue particle simulation toolkit on our GPU-accelerated HPC cluster. Microway allows you to quickly test your codes on the latest high-performance systems – you are free to upload and run your own software, although we also provide a variety of pre-compiled applications with built-in GPU acceleration. Our GPU Test Drive Cluster is a useful resource for benchmarking the faster performance which can be achieved with NVIDIA Tesla GPUs.

This post demonstrate HOOMD-blue, which comes out of the Glotzer group at the University of Michigan. HOOMD blue supports a wide variety of integrators and potentials, as well as the capability to scale runs up to thousands of GPU compute processors. We’ll demonstrate one server with dual NVIDIA® Tesla®  K40 GPUs delivering speedups over 13X!

Before continuing, please note that successful use of HOOMD-blue will require some familiarity with Python. However, you can reference their excellent Quick Start Tutorial. If you’re already familiar with a different software package, read through our list of pre-installed applications. There may be no need for you to learn a new tool.

Access a Tesla GPU-accelerated Compute Node

Getting started on our GPU system is fast and easy – complete this short form to sign up for HOOMD-blue benchmarking. We will send you an e-mail with a general list of commands when your request is accepted, but this post provides guidelines specific to HOOMD-blue tests.

First, you need SSH to access our GPU cluster. Don’t worry if you’re unfamiliar with SSH – we will send you step-by-step login instructions. Windows users have one extra step, but SSH is built-in on Linux and MacOS.

Run CPU and GPU-accelerated HOOMD-blue

Once you’re logged in, it’s easy to compare CPU and GPU performance: enter the HOOMD-blue directory and run the benchmark batch script which we have pre-written for you:

cd hoomd-blue
sbatch run-hoomd-on-TeslaK40.sh

Waiting for your HOOMD-blue job to finish

Our cluster uses SLURM to manage computational tasks. You should use the squeue command to check the status of your jobs. To watch as your job runs, use: watch squeue (hit CTRL+c to exit). Alternatively, the cluster can e-mail you when your job has finished if you update the HOOMD batch script file (although this must be done before submitting your job). Run:

nano run-hoomd-on-TeslaK40.sh

Within this file, add the following lines to the #SBATCH section (changing the e-mail address to your own):

#SBATCH --mail-user=yourname@example.com
#SBATCH --mail-type=END

If you would like to closely monitor the compute node which is executing your job, run squeue to check which compute node your job is running on. Log into that node via SSH and use one of the following tools to monitor the GPU and system status:

ssh node2
nvidia-smi
htop

(hit q to exit htop)

Check the speedup of HOOMD-blue on GPUs vs. CPUs

The results from the HOOMD-blue benchmark script will be placed in an output file named hoomd-K40.xxxx.output.log – below is a sample of the output running on CPUs:

======================================================
= Run CPU only lj_liquid_bmark_512K
======================================================
Average TPS: 21.90750

and with HOOMD-blue running on two GPUs (demonstrating a 13X speed-up):

======================================================
= Run Tesla_K40m GPU-accelerated lj_liquid_bmark_512K
======================================================
Average TPS: 290.27084

If you would like to examine the full execution sequence of a particular input, you will see that a log file has been created for each of the inputs (e.g., lj_liquid_bmark_512K.20_cpu_cores.output). If the HOOMD-blue job has any problems, the errors will be logged to the file hoomd-K40.xxxx.output.errors

The chart below shows the performance improvements for a CPU-only HOOMD-blue run (on two 10-core Ivy Bridge Intel Xeon CPUs) compared to a GPU-accelerated HOOMD-blue run (on two NVIDIA Tesla K40 GPUs):

Running your own HOOMD-blue inputs on GPUs

If you’re comfortable with shell scripts you can write your own batch script from scratch, but we recommend using the run-hoomd-your-files.sh file as a template when you’d like to try your own simulations. For most HOOMD-blue runs, the batch script will only reference a single Python script as input (e.g., the lj_liquid_bmark_512K.hoomd script). Reference the HOOMD-blue Quick Start Tutorial.

Once your script is in place in your hoomd-blue/ directory, you’ll need to ensure that the batch script is referencing the correct .hoomd input file. The relevant lines of the run-hoomd-your-files.sh file are:

echo "==============================================================="
echo "= Run CPU-only"
echo "==============================================================="

srun --mpi=pmi2 hoomd input_file.hoomd --mode=cpu > hoomd_output__cpu_run.txt
grep "Average TPS:" hoomd_output__cpu_run.txt

and for execution on GPUs:

echo "==============================================================="
echo "= Run GPU-Accelerated"
echo "==============================================================="

srun --mpi=pmi2 -n $GPUS_PER_NODE hoomd input_file.hoomd > hoomd_output__gpu_run.txt
grep "Average TPS:" hoomd_output__gpu_run.txt

As shown above, both the CPU and GPU runs use the same input file (input_file.hoomd). They will each save their output to a separate text file (hoomd_output__cpu_run.txt and hoomd_output__gpu_run.txt). The final line of each section uses the grep tool to print the performance of that run. HOOMD-blue typically measures performance in millions of particle time steps per second (TPS), where a higher number indicates better performance.

Will GPU acceleration speed up your research?

With our pre-configured GPU cluster, running HOOMD-blue across an HPC cluster isn’t much more difficult than running it on your own workstation. This makes it easy to compare HOOMD-blue simulations running on CPUs and GPUs. If you’d like to give it a try, contact one of our experts or sign up for a GPU Test Drive today!

Citation for HOOMD-blue:

Joshua A. Anderson, Chris D. Lorenz, and Alex Travesset – ‘General Purpose Molecular Dynamics Fully Implemented on Graphics Processing Units’, Journal of Computational Physics 227 (2008) 5342-5359
https://glotzerlab.engin.umich.edu/hoomd-blue/

Featured Illustration:

“Micellar crystals in solution from molecular dynamics simulations”, J. Chem. Phys. 128, 184906 (2008); DOI:10.1063/1.2913522
https://doi.org/10.1063/1.2913522

Citation for VMD:

Humphrey, W., Dalke, A. and Schulten, K., “VMD – Visual Molecular Dynamics” J. Molec. Graphics 1996, 14.1, 33-38
https://www.ks.uiuc.edu/Research/vmd/

The post Running GPU Benchmarks of HOOMD-blue on a Tesla K40 GPU-Accelerated Cluster appeared first on Microway.