A combination of array screening and geo-time metadata to investigate signs and origins of biological threats has been demonstrated under conditions of processing up to 800 terabytes within 24 hours.
Detection and judgment of origin in two stages
In order to compete with computing tools to investigate the origin of biological events from large quantities of genomic and environmental data, the U.S. Defense Advanced Research Program (DARPA) launched a virtual competition called the Bio-Attribution Challenge.
In the 1 phase, the ability to detect pathogens in complex environmental samples and reveal their characteristics was assessed. In the 2 phase, the task of finding the physical, chemical and design characteristics unique to the artificially designed pathogens and determining their origin was advanced.
Connect the characteristics of the array to location and time information.
In the final “Determination – Attribution” phase, detailed array screening and accurate geographical and time metadata had to be combined, not only with the characteristics that appear in the array, but also with the configuration to incorporate the corresponding location and time of the sample into the same analysis.
The sample information provided to participants in advance was minimized to near-real analysis conditions.The conditions for extracting the judgment material from genomic and environmental information were set without a detailed background explanation.
Process 600 to 800 terabytes within hours of 24
Participants in the final phase processed data sets from 600 to 800 terabytes within hours of 24. The evaluation subject was whether they could accurately identify indicators indicating the possibility that the pathogen was released or manipulated, not just the volume or speed of processing.
Under these conditions, the 8 team completed the final phase, which resulted in array screening and the process of connecting geo-time information as a computational task that handled hundreds of terabytes of data.
Analysis of simulated data in government management environments
The Lawrence Livermore National Laboratory (LLNL) created evaluation data modeling realistic and complex origin-specific scenarios.
The participant’s software was run in a safe environment under government control.DARPA has demonstrated that through this assessment, senior participants have demonstrated how to improve digitally on existing tools.