Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FlowCyt: A Comparative Study of Deep Learning Approaches for Multi-Class Classification in Flow Cytometry Benchmarking

Lorenzo Bini, Fatemeh Nassajian Mojarrad, Margarita Liarou, Thomas Matthes, Stéphane Marchand-Maillet

TL;DR

FlowCyt introduces the first public benchmark for multi-class single-cell classification in flow cytometry, providing a richly annotated bone marrow dataset from 30 patients with 12 markers and ground-truth labels for five hematologic cell types. The authors compare traditional baselines with graph neural networks, showing that graph-encoded data and models like GAT yield superior performance in both inductive and semi-supervised transductive settings, especially as dataset size scales to up to one million cells per patient. Beyond classification, FlowCyt supports sub-population and total-population tasks and facilitates clustering, dimensionality reduction, and trajectory inference, demonstrating its versatility for hematology and immunology analyses. By offering standardized evaluation and diverse analytical capabilities, FlowCyt aims to accelerate method development and clinical translation for automated single-cell analysis.

Abstract

This paper presents FlowCyt, the first comprehensive benchmark for multi-class single-cell classification in flow cytometry data. The dataset comprises bone marrow samples from 30 patients, with each cell characterized by twelve markers. Ground truth labels identify five hematological cell types: T lymphocytes, B lymphocytes, Monocytes, Mast cells, and Hematopoietic Stem/Progenitor Cells (HSPCs). Experiments utilize supervised inductive learning and semi-supervised transductive learning on up to 1 million cells per patient. Baseline methods include Gaussian Mixture Models, XGBoost, Random Forests, Deep Neural Networks, and Graph Neural Networks (GNNs). GNNs demonstrate superior performance by exploiting spatial relationships in graph-encoded data. The benchmark allows standardized evaluation of clinically relevant classification tasks, along with exploratory analyses to gain insights into hematological cell phenotypes. This represents the first public flow cytometry benchmark with a richly annotated, heterogeneous dataset. It will empower the development and rigorous assessment of novel methodologies for single-cell analysis.

FlowCyt: A Comparative Study of Deep Learning Approaches for Multi-Class Classification in Flow Cytometry Benchmarking

TL;DR

FlowCyt introduces the first public benchmark for multi-class single-cell classification in flow cytometry, providing a richly annotated bone marrow dataset from 30 patients with 12 markers and ground-truth labels for five hematologic cell types. The authors compare traditional baselines with graph neural networks, showing that graph-encoded data and models like GAT yield superior performance in both inductive and semi-supervised transductive settings, especially as dataset size scales to up to one million cells per patient. Beyond classification, FlowCyt supports sub-population and total-population tasks and facilitates clustering, dimensionality reduction, and trajectory inference, demonstrating its versatility for hematology and immunology analyses. By offering standardized evaluation and diverse analytical capabilities, FlowCyt aims to accelerate method development and clinical translation for automated single-cell analysis.

Abstract

This paper presents FlowCyt, the first comprehensive benchmark for multi-class single-cell classification in flow cytometry data. The dataset comprises bone marrow samples from 30 patients, with each cell characterized by twelve markers. Ground truth labels identify five hematological cell types: T lymphocytes, B lymphocytes, Monocytes, Mast cells, and Hematopoietic Stem/Progenitor Cells (HSPCs). Experiments utilize supervised inductive learning and semi-supervised transductive learning on up to 1 million cells per patient. Baseline methods include Gaussian Mixture Models, XGBoost, Random Forests, Deep Neural Networks, and Graph Neural Networks (GNNs). GNNs demonstrate superior performance by exploiting spatial relationships in graph-encoded data. The benchmark allows standardized evaluation of clinically relevant classification tasks, along with exploratory analyses to gain insights into hematological cell phenotypes. This represents the first public flow cytometry benchmark with a richly annotated, heterogeneous dataset. It will empower the development and rigorous assessment of novel methodologies for single-cell analysis.
Paper Structure (27 sections, 1 equation, 3 figures, 7 tables)

This paper contains 27 sections, 1 equation, 3 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Flow Cytometry
  3. Manual FC Data Analysis
  4. Related Work
  5. Dataset Description
  6. Biological Description
  7. Technical Description
  8. Data and labels:
  9. Files:
  10. Structure of the Benchmark
  11. Classification Task:
  12. Sub-Population Classification:
  13. Total Population Classification:
  14. Other Tasks:
  15. Experiments
  16. ...and 12 more sections

Figures (3)

  • Figure 1: The dataset was split into seven segments for analysis. Six of these segments were used as training data, while one was reserved as test data. Additionally, 10% of each training segment was randomly selected to create a validation set.
  • Figure 2: Feature importance as highlighted by the model. For the sake of simplicity, we remind the reader to match the labels with the corresponding Table \ref{['tab:1']} for future explanations.
  • Figure 3: t-SNE projection for one random patient, for the transductive learning task.