Utilization of High Performance Computing for High Throughput Identification of Human Biosignatures Derived From Image Analytics of Chemical and Bio

Project: Research project

Project Details


The INSIGHTS project is an experimental research project led by Dr. Thomas Lamkin of AFRL/RHXBC that aims to discover biological signatures and effects on such biosignatures resulting from chemical and genetic interrogations (referred to as treatments) that are beneficial for human performance, medical intelligence, or therapeutic applications. High-throughput, high-content biological screening is performed to generate>12,000 images per experimental plate. These Images are analyzed using a code called AIME_BIOS to measure values for>11,000 features for each cell identified within the images. Data mining techniques are used to identify features that can be best used to discriminate between healthy and infected cells, classify all identified cells, and determine treatments of interest. Due to the large computational and data requirements, the use of high performance computing (HPC) is the only available mechanism to efficiently identify the treatments of interest. The current INSIGHTS computational pipeline uses programming languages that are not well suited to HPC. AIME_BIOS is implemented in C#; feature selection, cell classification, and hit selection algorithms are implemented in R and Java. It is estimated that the total time to solution could be reduced by an order of magnitude by porting the algorithms to languages suitable for HPC. This proposed pre-planned effort has three objectives: (a) port selected AIME_BIOS algorithms from C# to C++, (b) develop an MPI-based modular framework for the C++ AIME_BIOS, and (c) develop and/or leverage previously developed (possibly parallel) C/C++ implementations of feature selection, cell classification, and hit selection algorithms.
Effective start/end date3/5/149/30/14


  • US General Services Administration (GSA): $30,000.00


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.