Workstation for sequence data analysis/machine learning (optimized for system requirements) | TEGAKARI, an information dissemination media for research and development

Case No. PC-25000268A customer who visited our website inquired about a workstation that could handle both nanopore sequencing data analysis and machine learning.
We plan to perform epigenetic analysis, de novo assembly, RNA-Seq analysis, metagenomic analysis, and other processing, and will use the following software:

General-purpose analysis and ML infrastructure: Samtools, XGBoost, LightGBM
Genome assembly and evaluation: SPAdes, BUSCO, Prodigal, FastANI, GTDB-Tk
Nanopore base calling: Guppy, Dorado, Remora, DeepMod2

The OS will be Ubuntu, and I would like a quote including installation work within a budget of 400 million yen.

Based on the information you provided, we proposed the following configuration.

CPU	AMD EPYC 9554 3.1GHz 64C/128T
memory	Total 512GB DDR5 5600 REG ECC 64GB x 8
Storage 1	Two 4TB SSD M.2 NVMe Gen5
Storage 2	24TB HDD S-ATA
Video	NVIDIA RTX PRO 6000 Max-Q 96GB
network	on board (10GBase-T x2)
Housing + power supply	Mid-tower 1500W 80PLUS PLATINUM
OS	Ubuntu 24.04
Other	On-site installation work (unpacking, installation, cabling, startup check)

Large memory configuration that prioritizes tool requirements

This configuration is optimized based on the system requirements. It prioritizes the number of CPU cores and memory capacity, especially for large assemblies and batch processing of many samples. The EPYC 9554 (64 cores) and 768GB of memory provide advantages for memory-intensive analyses.

Case No. PC-25000268While upgrading the system, the configuration emphasizes versatility and can also handle machine learning.Case No.PC- TW3J253928 .

About CPU

The AMD EPYC 9554's 64-core configuration provides high parallel processing performance, making it ideal for workloads that require parallel processing of large numbers of samples. It demonstrates efficiency in environments where multiple jobs are run simultaneously, or in genome analysis tools that require high CPU resources.

Memory capacity

Among the software you plan to use, there are several tools that recommend a memory capacity of around 256GB to 512GB. Therefore, taking into account the scale of analysis and future data expansion, we have installed a generous 768GB. This will also contribute to improving the stability of long-term jobs.

GPU selection

This configuration is equipped with an NVIDIA RTX 6000 Ada (Max-Q 96GB) to support machine learning processing for research purposes.
It has high compatibility with deep learning frameworks and is highly practical for model learning and inference processing.

Dorado is an example of a high-end GPU that assumes double-precision calculations when using GPU acceleration.
Therefore, rather than simply configuring a system that meets Dorado's high-precision GPU requirements, we decided to use the RTX series as a realistic and effective option that can flexibly accommodate a wide range of research applications, particularly machine learning.

The RTX series is well-suited for deep learning frameworks and is easily adaptable to a wide range of research tasks. Its configuration allows for flexible adaptation to future expansion and workflow evolution.

Reference: Dorado GPU requirements

Reference: Benchmarking the Oxford Nanopore Technologies basecallers on AWS

Storage and Operability

For workflows requiring high-speed access, the system is equipped with two 4TB NVMe Gen5 SSDs for high-speed temporary data processing.
In addition, by using a 24TB HDD, the storage configuration is suitable for long-term storage and reanalysis of sequence data.
Furthermore, we also handle on-site installation (unpacking, setting up, and startup check) so that you can begin research immediately after installation.

Optimization in line with operational requirements

It was designed to meet the delivery deadline and budgetary requirements while also meeting the desired specifications. Its strength is its ease of maintaining a stable analysis pipeline in a research environment where a variety of tools are used.

Tegara's custom-made PC production service not only caters to initial use, but also supports system expansion in anticipation of future expansion of research scale.
We not only propose configurations that meet various software requirements, but also accept consultations regarding the construction of an entire research environment.
Please feel free to contact us and we will provide the best solution to suit your needs.

Keyword

・What is Samtools?
Samtools is a set of standard tools widely used for processing and analyzing next-generation sequencing (NGS) data. It enables high-speed processing of alignment data such as BAM and CRAM, including sorting, indexing, filtering, and visualization, and is widely used as the foundation of genome analysis workflows.

Reference: GitHub – samtools/samtools: Tools (written in C using htslib) for manipulating next-generation sequencing data

・What is XGBoost?
XGBoost is a machine learning framework that implements the gradient boosting decision tree (GBDT) algorithm with high speed and accuracy. It has a wide range of applications including classification, regression, and ranking, and is also widely used in the bioinformatics field, including genome analysis, disease risk prediction, and gene expression analysis.

https://xgboost.ai/Reference: XGBoost Official Site

・What is LightGBM?
LightGBM is a fast and memory-efficient gradient boosting framework developed by Microsoft. It is suitable for analyzing large amounts of data and is used for SNP analysis, methylation data classification, disease prediction models, and more.

Reference: LightGBM Documentation

What is SPAdes?
SPAdes is a de novo assembler widely used for microbial genome assembly. It supports hybrid assembly integrating multiple platforms, including Illumina, Nanopore, and PacBio, and is capable of constructing high-quality contigs.

Reference: SPAdes – Genome Assembly Toolkit

・What is BUSCO?
BUSCO is a representative tool for assessing the completeness of genome assemblies. It uses a highly conserved set of single-copy genes to objectively assess assembly quality.

References: Release · ezlab / busco · GitLab

・What is Prodigal?
Prodigal is a gene prediction tool specialized for microbial genomes. It is capable of extracting prokaryotic gene regions with high accuracy and also performs well in metagenomic analysis.