Home NVIDIA GPU Powers Qitan Tech to Significantly Accelerate Sequencing Software Data Processing

NVIDIA GPU Powers Qitan Tech to Significantly Accelerate Sequencing Software Data Processing

Dec 01, 2020 14:01 CST Updated 14:01
NVIDIA

Artificial Intelligence Computing Service Provider

Case Brief

In this case, by leveraging NVIDIA GPUs, Qitan Technology’s nanopore gene sequencer achieved a 20% increase in sequencing accuracy and a 10-fold boost in sequencing speed, eliminating the need for additional feature extraction processes. Meanwhile, the NVIDIA CUDA parallel computing platform ensured the smooth execution of hyperparameter tuning and large-scale data training.


This case primarily utilizes NVIDIA GPUs and the CUDA computing platform.


Client Profile and Application Background

Qitan Technology is dedicated to the independent research, development, manufacturing, and application of nanopore gene sequencers along with their supporting reagents and consumables. It is the first and currently the only company in China that has successfully developed a nanopore gene sequencer and obtained valid sequencing data. The company completed the development of its proprietary proof-of-concept prototype and engineering prototype of the nanopore gene sequencer in 2017 and 2018, respectively.


Nanopore gene sequencing technology represents the world’s most advanced fourth-generation sequencing platform. Compared with the currently dominant second-generation technologies, it offers advantages such as longer read lengths, faster sequencing speeds, lower per-run costs, and compact, portable instrumentation, thereby enabling effective complementarity with second-generation technologies in specific application niches. As a high-tech in vitro diagnostic product, the accuracy of test results and the speed of detection are key performance indicators for nanopore gene sequencers.


Customer Challenges and Application Solutions

An ideal gene sequencer should deliver highly accurate detection results in the shortest possible time, providing robust data support for scientific inference and clinical diagnosis. Qitan Technology leverages neural networks to analyze sequencing waveform data, employing NVIDIA GPUs in conjunction with the QuartzNet network architecture and a CTC decoder. This approach eliminates the need for additional feature extraction processes, significantly enhancing sequencing accuracy by 20% and increasing sequencing speed tenfold. These advancements enable broader applications in emerging fields such as pathogen detection, food safety, and public health.

 

In the ongoing effort to enhance the accuracy and sequencing speed of gene sequencers, Qitan Technology has employed methods such as dataset expansion and neural network hyperparameter tuningAlgorithm optimization was performed by leveraging the large memory capacity and multi-core computational advantages of NVIDIA GPUs, significantly reducing computation time while ensuring accuracy.


As is well known, training large-scale neural networks with big data is extremely computationally intensive, often requiring weeks to complete. Meanwhile, hyperparameter tuning necessitates iterative parameter adjustments, with each update requiring small-scale training to validate performance. To address these challenges, Qitan Technology leverages NVIDIA’s CUDA parallel computing platform to ensure the smooth execution of both hyperparameter tuning and large-scale data training.


Efficacy and Impact

NVIDIA GPUs have significantly accelerated the data processing speed of Qitan Technologies’ sequencing software. Leveraging the robust computational power provided by GPUs, Qitan Technologies’ nanopore gene sequencer can perform real-time analysis of sequencing waveforms and output results in real time. The massive number of streaming processors in GPUs makes seq2seq tasks based on convolutional neural networks feasible. A GPU with 1,000 streaming processors can achieve a sequencing throughput of approximately 100 MB/h, whereas an 8-core CPU can only manage 10 MB/h.

 

“NVIDIA’s products have provided strong technical support for Qitan Technology’s development of nanopore gene sequencers, enabling the company to achieve one technological breakthrough after another with speed and stability. With significant advantages including longer read lengths, faster speeds, lower costs, and a compact, portable design, nanopore gene sequencers are poised to drive the deeper integration of gene sequencing services into scientific research, clinical medicine, and consumer applications. Through our collaboration with NVIDIA, Qitan Technology is more confident than ever in developing nanopore gene sequencers that rival the most advanced international standards,” said Dr. Hu Geng, CEO of Qitan Technology.