Using computational science to enable and accelerate research

The Department of IT Research Computing at Roswell Park provides training and expertise in a variety of scientific software codes, programming languages, computational methods, visualization platforms, machine learning tools, computational pipelines and software workflows.

We also offer training and access to high performance computing facilities maintained by Roswell Park's IT department, as well as our partners at the University at Buffalo Center for Computational Research (UB CCR).

Computing assets

Roswell Park HPC Cluster

Roswell Park HPC cluster

1,600 processors (32 TFLOPS)
450 TB storage, 10.5 TB RAM

UB CCR Cluster

UB CCR cluster

24,320 processors (1.2 PFLOPS)
4.7 PB storage, 135 TB RAM

Training topics and expertise

  • Quarterly “on-boarding” of new high-performance computing (HPC) users
  • Machine learning techniques and related software, including Keras, TensorFlow, PyTorch
  • Introductory programming, including Python, R, MATLAB, C/C++, or Fortran
  • Using cloud computing services, including configuring OpenStack and AWS instances, using elastic storage, cloud service cost models
  • Using containers, including Docker and Singularity, creating containers, running containerized software
  • Advanced programming topics, including high-performance computing, data ingestion, data modeling and regression, plotting and visualization, computational pipelines and software workflows
  • Scientific software, including ParaView, VTK, MeshLab, ABAQUS, ANSYS, COMSOL, LS-DYNA, StarCCM, OpenFOAM, OpenSEES, PETSc, LAMMPS and NAMD

Grant support

IT Research Computing staff can serve as co-PI or Senior Personnel on grant proposals developed by Roswell Park researchers. For a given grant, up to 10% effort (4 hours per week) per staff member may be devoted to the following grant-related activities:

  • Pre-award phase (proposal development): Facilities descriptions; descriptions of software workflows, architectures, and/or pipelines; budgets and milestones for software development; data management plans
  • Award phase (direct contributions to funded grant): Progress reports; student training and mentorship; developing and implementing computational algorithms; integrating software tools into new scientific workflows
  • Post-award phase (reporting results): Project summaries; manuscript preparation, submission, and revision; archiving data; maintaining software repositories

Cost recovery is required for direct contributions of IT staff to funded grants – be sure to budget appropriately!

Research highlights

Data for a research study on breast cancer

Studies of Breast Cancer in African-Americans
Principal Investigator: Dr. Song Yao

Funded by the National Cancer Institute, Dr. Yao’s research seeks to: (1) explain why aggressive forms of breast cancer are more common in African-American women, (2) learn how cancer metastasizes and develops resistance to treatment, (3) explore opportunities to enhance patient recovery following surgery, and (4) accelerate the development of new immunotherapies.

Accelerating Immune-base Cancer Treatment

Accelerating Immune-base Cancer Treatment
Principal Investigator: Dr. Alan Hutson

Dr. Hutson’s research group is leveraging high performance computing to accelerate the development of immune-based cancer treatment approaches and prevention strategies. The research supports the Immuno-Oncology Translational Network (IOTN) – a consortium of research centers that are participating in the Cancer Moonshot program sponsored by the National Cancer Institute.

A graph showing the outcomes of radiation

Optimizing Radiation Treatment Plans
Principal Investigator: Dr. Daryl Nazareth

Dr. Nazareth’s group has been using the UB CCR supercomputer to develop optimized radiation treatment plans for cancer patients. Using a technique known as simulation-based optimization, the supercomputer identifies treatment plans that maximize the radiation dose delivered to the tumor region while simultaneously minimizing the dose absorbed by healthy tissue.

An illustration of a cross section of a radiation burn

Automatic Classification of Radiation Burns
Principal Investigator: Dr. Simon Fung-Kee-Fung

Dr. Fung-Kee-Fung and his team are developing a machine learning model for classifying the severity of radiation burns that are a common byproduct of radiation therapy. Initial datasets have focused on breast cancer patients and have achieved 85% classification accuracy.

Imagine of a patient's head and neck area

Automatic Segmentation of Head and Neck Cancers
Principal Investigators: Dr. Anh Le and Dr. John Asbach

This research is using an adaptation of the U-Net neural network model architecture for automatic OAR (organ-at-risk) segmentation of medical images. The current focus is on the automatic segmentation of head and neck cancers.

Flow Cytometry

Advanced Visualization of Flow Cytometry Data
Principal Investigators: Flow Cytometry Team

The Flow Cytometry Shared Resource is accelerating its workflow by using “t-SNE” (t-distributed stochastic neighbor embedding) - an unsupervised machine learning algorithm for embedding high-dimensional data for visualization in a low-dimensional space of two or three dimensions.

Submitting help requests through ServiceNow

NOTE: For internal teams only

Request IT

  • Obtain an account on the HPC cluster
  • Install scientific software
  • Set up group or individual training
  • Assess computing needs of a research project
  • Make research-related software and hardware purchases
  • Port, profile, and optimize research codes

Get Started