White Papers

ManualsBrandsDell ManualsConverged InfrastructureHigh Performance Computing Solution Resources

DELL EMC POWER EDGE

R940 MAKES DE

NOVO ASSEMBLY EASIER

Genome Assembly on Deep Sequencing data with

SOAPdenovo2

ABSTRACT

De novo assemblies are memory intensive since the assembly algorithms need to

compare every read with every other read imposing a time complexity of O(n

). The

memory requirement has been growing as more sequence reads are being generated

than ever before, so a big memory machine like the Dell EMC PowerEdge

R940 is

required to assemble deep sequencing data. Here we introduce R940 with Intel

Xeon

Scalable Processor, which reduces the running time by 27% and consumes 39.5% less

memory on SOAPdenovo2 to assemble 3.2 billion reads, in comparison to its

predecessor, R930.

July, 2017

Summary of content (10 pages)

PAGE 1
DELL EMC POWER EDGE® R940 MAKES DE NOVO ASSEMBLY EASIER Genome Assembly on Deep Sequencing data with SOAPdenovo2 ABSTRACT De novo assemblies are memory intensive since the assembly algorithms need to compare every read with every other read imposing a time complexity of O(n2). The memory requirement has been growing as more sequence reads are being generated than ever before, so a big memory machine like the Dell EMC PowerEdge® R940 is required to assemble deep sequencing data.
PAGE 2
TABLE OF CONTENTS EXECUTIVE SUMMARY ...........................................................................................................3 AUDIENCE ........................................................................................................................................ 3 INTRODUCTION ........................................................................................................................4 De novo assembly? .................................................................
PAGE 3
EXECUTIVE SUMMARY In the last decade, modern society has continued to improve the quality of life through better healthcare, producing and consuming sustainable food and energy, and protecting our environment. All of these societal advancements are tightly related to progress in the life science domain. Bioinformatics emerged from the massive amount of data that is now available in these fields and the advancement of computer technology takes a key role to finding solutions.
PAGE 4
INTRODUCTION De novo assembly? Genome assembly programs perform complex computations to join billions of fragmented reads of DNA. Short-read sequencing (Next Generation Sequencing (NGS)) technologies turn the haploid human genome into 2- to 3-billion pieces of jigsaw puzzle with 100 copies of each piece (1). This puzzle is complicated further by some pieces being missing or containing errors.
PAGE 5
Memory capacity limitations on a shared memory architecture Modern CPU micro-architecture supports multiple independent channels to the DIMMs through integrated memory controllers (IMCs) (5). Both Intel® Haswell and Broadwell support quadruple memory channels with two IMCs per processor (6) (7). These multiple independent channels enable concurrent access to multiple DIMMs and distributes the data amongst the DIMMs providing more bandwidth for accessing the data.
PAGE 6
Test Configurations and Data Since the main interest of this study is to make a reasonable observation for memory consumption behavior between two systems, all DIMM slots in both systems were fully populated to match the amount of memory to 1.5 TB. Although other components can influence the benchmark results like storage configuration, those influences should be marginal since the performance of de novo assembly applications is largely bounded by CPU/memory.
PAGE 7
In server to server comparison, R940 with 110 cores reduced running time by 37% for the tested configurations (Figure 4). Figure 4 Running time comparisons between R930 and R940 as in Table 1: 2 free cores were reserved for both systems to avoid overstressing the systems. Also, the scaling over various input sizes with a fixed number of cores shows nearly linear behavior on R940 as shown in Figure 5.
PAGE 8
Peak Memory Consumption The improvement in peak memory consumption is unusual given that an application typically uses a fixed amount of memory according to input data size and/or the number of threads allocated. However, SOAPdenovo2 consumes 39.5% less memory on PowerEdge R940 than PowerEdge R930 for 3.2 billion read data. Figure 6 shows the peak memory consumption over various data sizes with 62 cores on both systems.
PAGE 9
Figure 8 ScanRead tests by pmbw suit Figure 9 ScanWrite tests by pmbw suit Conclusion The Dell EMC 14th Generation server, PowerEdge R940, is superior to its predecessor in several ways. By taking advantage of more memory channels supported with Skylake-SP and faster memory speed, R940 demonstrates speed gain as well as more efficient memory usage. SOAPdenvo2 actually consumes less memory since it process more data faster on R940. This means that even 6 billion reads data can be processed on R940 with 1.
PAGE 10
References 1. What is Next-Generation DNA Sequencing. [Online] https://www.ebi.ac.uk/training/online/course/ebi-next-generation-sequencingpractical-course/what-you-will-learn/what-next-generation-dna-. 2. Sanger Sequencing. [Online] https://en.wikipedia.org/wiki/Sanger_sequencing. 3. De novo genome assembly: what every biologist should know. Baker, Monya. 2012, NATURE METHODS, pp. 333–337. 4. De Bruijn graph. [Online] https://en.wikipedia.org/wiki/De_Bruijn_graph. 5. Multi-channel memory architecture.