BL-FlowSOM: Consistent and Highly Accelerated FlowSOM Based on Parallelized Batch Learning.

Journal: Cytometry. Part A : the journal of the International Society for Analytical Cytology
Published Date:

Abstract

The recent increase in the dimensionality of cytometry data has led to the development of various computational analysis methods. FlowSOM is one of the best-performing clustering methods but has room for improvement in terms of the consistency and speed of the clustering process. Here, we introduce Batch Learning FlowSOM (BL-FlowSOM), which is a consistent and highly accelerated FlowSOM based on parallelized batch learning. The change of the learning algorithm from online learning to batch learning with principal component analysis initialization improves consistency and eliminates randomness in the clustering process. It also enables the parallelization of the learning process, leading to significant acceleration of the clustering process with clustering quality equivalent to that of FlowSOM. BL-FlowSOM is available on Sony's Spectral Flow Analysis (SFA)-Life sciences Cloud Platform (https://www.sonybiotechnology.com/us/instruments/sfa-cloud-platform/).

Authors

  • Fumitaka Otsuka
    Life Science Technology Research & Development Department, Technology Development Laboratories, Sony Corporation, Tokyo, Japan.
  • Kenji Yamane
    Life Science Technology Research & Development Department, Technology Development Laboratories, Sony Corporation, Tokyo, Japan.
  • Koji Futamura
    Technology & Engineering Center, Sony Corporation, Yokohama, Japan.
  • Junichiro Enoki
    Technology & Engineering Center, Sony Corporation, Yokohama, Japan.
  • Yuji Nishimaki
    Technology Infrastructure Center, Sony Group Corporation, Tokyo, Japan.
  • Yoshiki Tanaka
    Chukyo Medical Co. Inc., Nagoya, Japan.
  • Akihide Higuchi
    Life Science & Technology Business Unit, Sony Corporation, Yokohama, Japan.
  • Motohiro Furuki
    Life Science Technology Research & Development Department, Technology Development Laboratories, Sony Corporation, Tokyo, Japan.