PEAKS Online, means high-throughput proteomics data analysis for multiple users on a network. Although the term ‘Online’ is often associated with the world wide web, PEAKS Online should not be confused with public access. Instead, PEAKS Online is a software package which includes a Server license and Client licenses and can be hosted on a public cloud such as AWS or Google Cloud or it can be deployed on a private network such as your own high-performance computing cluster. This cloud-based architecture allows PEAKS Online to be fully parallelized and scalable to your lab’s needs and is ideal for processing large-scale projects.
In PEAKS Online X dual interfaces are available to ensure easy integration into any proteomics workflow. The Web Client Interface allows users to setup and submit projects visually, as well as review and validate their results. The Client CLI on the other hand can be integrated into existing pipelines to continue automated data processing.
Peaks Online X+ is designed to facilitate accurate and sensitive proteomic analysis using the PEAKS X+ software workflows, but with higher performance on a shared resource. Users will be able to perform De Novo Sequencing , PEAKS DB , PEAKS PTM , SPIDER, as in traditional PEAKS Studio package. Optional PEAKS Q , PEAKS IMS, modules can be enabled, for labelled and label free protein quantification and ion mobility data support respectively.
Like PEAKS Studio, raw data from the mass spectrometer does not need to be converted prior to adding the data into the PEAKS Online X workflow. Data can be added to PEAKS Online X projects either from the local file system (through the browser or command line client), or, remote data repositories (for cases where your data is not stored locally). This makes it easy to submit projects from any computer with access to your network, even if it does not have direct access to your data file storage.
Once data has been uploaded onto the PEAKS Online X server, setup a PEAKS workflow or select from a predefined list. With PEAKS Online X, users can define standardized workflows to implement protocols in the lab for regulatory compliance and establish project specific approaches that can be used across the whole group.
However, if an analysis needs to be rerun with a new set of parameters, the whole workflow does not need to be reprocessed. Using the PEAKS Online X technology, samples can be added/removed to a project and workflows can be modified “on the fly”. PEAKS Online X will only process the information that needs to be updated while keeping any of the relevant information generated previously.
Once the desired results have been generated, users can export the results for each individual result node, like in PEAKS Studio, or export all the results at once in a single step.
PEAKS Online X uses the latest technology of distributed computing to fully utilize the computing power of your hardware. The PEAKS Online X architecture is built on the highly popular Apache Cassandra database system, which is used in heavy-load applications like Facebook and Netflix. PEAKS Online X further distributes its computational workload to different workers using the Akka Actor system. This new setup gives PEAKS Online X the ability to do a lot of things that is was not feasible in the PEAKS Studio desktop version. In particular, PEAKS Online X is easily able to take on large-scale projects by more efficiently handling the increased workload of large cohort proteomics studies. In addition, its ready-to-scale, high-performance setup where the throughput and performance can be adjusted dynamically by changing the hardware configuration. Given the right resources, PEAKS Online X can speed up data processing at least 10 times faster than PEAKS Studio and can handle 1000 samples or more.
In a recent benchmarking study, we compared the scalability of PEAKS Online X as we increase the number of computing resources. The testing data set contained 56 samples where each sample was made up of 12 fractions. In total, we had 672 three-hour MS runs which composed of 5 million MS1 scans and 30 million MS2 scans. With the standard PEAKS Online X 32-threads license, it took about 10 days to finish the project, from data loading , data refinement, de novo sequencing, PEAKS DB, PEAKS PTM and SPIDER. However, when we increased the CPU cores, the performance increase linearly. With 512 cores, it took a little more than half day to finish the whole analysis.
PEAKS Online is distributed as a software package that includes a Server license with multiple Client access. The base package of PEAKS Online includes 4 Client licenses and can utilize up to 32 logical cores/threads. However, PEAKS Online is ready to scale, and researchers can increase the performance and/or number of users to meet the needs of any research group. The strength of the PEAKS Online Server and number of Client licenses can be purchased to align the ideal solution.
*Note: The number of cores/threads indicates the number of usable cores or threads that can be allocated to the PEAKS Online Server’s worker nodes.
Users can purchase additional individual client licenses or a site client license to meet their lab’s needs. For quantification and ion mobility data support, purchase the optional PEAKS Q and PEAKS IMS add-on respectively.
Overall, PEAKS Online X is composed of 3 components: the Database Node(s), Master Node, and Worker Node(s). These can be distributed to different computers, or placed on the same machine. Each component has a unique pattern of computing resource usage and hardware requirements.
Database Node(s) of PEAKS Online X store all application data and are the base for all proteomics processing. Since PEAKS Online X is a distributed computing framework that can run from multiple machines, we use the popular distributed database system Cassandra as the main data storage to provide I/O performance at scale. Each data node, as part of a Cassandra cluster, has high demand for memory space and disk I/O speed.
The Master Node is the central hub of PEAKS Online X’s computing framework. It takes charge of scheduling, dispatching, and synchronization of computing tasks. Although it does not perform any data processing, it takes care of the web based user interface, loading of raw data and exporting of result data, and would benefit more from high performance CPUs.
Worker Nodes are responsible for the actual data processing and computation. PEAKS Online X is easily scalable, so a worker node can be configured to use a customized number of CPU threads. As a rule of thumb, a worker node needs 2GB available memory for each computing thread and another 2GB spare memory for its own usage. Other than a few GB for logging, a worker node generally has no requirement for hard disk I/O speed nor space.
|32 Threads Server License||128 Threads Server License|
|Master||16 Threads, 20 GB Memory||16 Threads, 32 GB Memory|
|Cassandra||8 Threads, 16 GB Memory, SSD (5T+)||8 Threads Each, 16 GB Memory Each, SSD (5T+)|
|Workers||(4x): 32 Threads Total (8 Thread Per), 80 GB Memory (20 GB Per)||(16x): 128 Threads Total (8 Thread Per), 320 GB Memory (20GB Per)|
|Suggested Configuration||All-in one Installation:|
INTEL XEON GOLD 5120 Processor X2
(56 Threads including hyper threading)
Memory: 144 GB
SSD: 500G (for OS)
SSD: 5T (for Cassandra)
|Cluster with 2 machines, each has the following hardware:
INTEL XEON GOLD 6238 Processor X2
(88 threads including hyper threading)
Memory: 256 GB
SSD: 500G (for OS)
SSD: 5T (for Cassandra)
Note: Although PEAKS Online X is best used in a cluster setting with dozens to hundreds of nodes, if you only have limited computing resources, one powerful workstation/server, you can still use it in standalone mode. In fact, all components of PEAKS Online can be started from the same machine. But in this case, you must make sure you have enough computing resources to accommodate the needs from all components
Note: Cassandra has very intense usage of the hard drive. If everything is installed on the same machine, it’s highly recommended to have a separate high speed hard drive specifically for the data component.
Lightbend, Inc. (2011-2019). Akka: Build powerful reactive, concurrent, and distributed applications more easily. https://akka.io/.
The Apache Software Foundation (2016). Apache Cassandra. http://cassandra.apache.org/.
Xin, L. et al. Towards Real-time Proteomics Data Analysis using PEAKS Online 8.5. ASMS. ThP 461. 7/6/2018