Automated Data Analysis Pipelines that Drop-in Replicate and Extend Manual Analysis

Automated Data Analysis Pipelines that Drop-in Replicate and Extend Manual Analysis

Automated Data Analysis Pipelines that Drop-in Replicate and Extend Manual Analysis (24031)

Ryan Brinkman ¹

BC Cancer Agency, Vancouver, BC, Canada

Conventional manual data analysis cannot provide a complete analysis of datasets generated by the current generation of flow cytometry instruments generating 50 parameter data on each of hundreds of thousands of single cells for each of up to thousands of samples in study. Algorithms have reached a level of maturity that enables them to match and in many cases exceed the results produced by human experts [1,2,3]. An overview will be provided of robust, reproducible and rapid data analysis pipelines for both automatic identification of cell populations of known importance (e.g., diagnosis by identification of pre-defined cell population) and for exploratory analysis of cohorts (e.g., discovery of cell populations that correlate with patient subgroups). Real-word examples of how automated discovery and diagnosis approaches have been used in basic and clinical research will be used illustrate the power of these approaches in practice.

Aghaeepour et al., Critical assessment of automated flow cytometry data analysis techniques. Nature Methods. (2013) PMC3906045.
Aghaeepour et al., A benchmark for evaluation of algorithms for identification of cellular correlates of clinical outcomes. Cytometry Part A (2015) PMC4874734.
Finak et al., Standardizing Flow Cytometry Immunophenotyping Analysis from the Human ImmunoPhenotyping Consortium. Nature Scientific Reports (2016). PMC4748244

#ACS2016

Authors contributing to this abstract.

Brinkman, R

My group is focused on applying bioinformatics techniques to flow cytometry data. Flow cytometry is a technique that is widely used within the biomedical community. New high throughput methods can generate up to a thousand flow cytometry data files per day and each data file can consist of millions of multiparametric descriptions of individual cells. However flow cytometry lacks data standards for fully representing both the methods used and the data generated, hindering the handling, exchange and dissemination of scientific research. Consequently, there are a variety of challenges to archiving, analyzing and reporting the results of high throughput flow cytometry experiments. Furthermore, only rudimentary bioinformatics tools exist to manage and mine high throughput data. We are leading an international effort to develop a systemic approach to modeling, capturing, analyzing and disseminating flow cytometry data. Initial work is focused on developing a community-based standard for recording and reporting flow cytometry data. Software implementations of this standard (UML, XML, SQL, Java) and an ontology (Ontology for Biomedical Investigations) being created to facilitate data exchange between both software components and collaborative groups. Statistics packages (e.g., flowCore/flowUtils/flowClust) are also being developed to allow users to implement analyses in a high throughput fashion, as well as exchange these analyses in more meaningful ways then are currently available. We are also testing our high throughput analysis methods to analyze flow cytometry data on datasets from the British Columbia Cancer Agency and the British Children's Hospital including developing methods for the high throughput analysis of lymphoma, Graft versus Host Disease and innate immunity. Some of these data sets are also available as part of the flowCAP project. We are also developing user-friendly software that implements some of the new data analysis and visualization methods we and others have developed. My research is supported by the NIH, CIHR, Genome Canada, Genome BC, The Michael Smith Foundation for Health Research and the BC Cancer Agency.

Automated Data Analysis Pipelines that Drop-in Replicate and Extend Manual Analysis (24031)

Add notes

Brinkman, R

Login