(362b) Characterizing Host Immune Cell and Local Transcriptional Dynamics during Influenza Infection | AIChE

(362b) Characterizing Host Immune Cell and Local Transcriptional Dynamics during Influenza Infection


Wang, M. - Presenter, University of Pittsburgh
Fukuyama, S., University of Tokyo
Kawaoka, Y., University of Wisconsin-Madison
Shoemaker, J. E., University of Pittsburgh

Influenza virus is a threat to human health, which results in millions of cases, hundreds of thousands of hospitalizations, and tens of thousands of deaths every year in the United States [1]. Several aspects of the immune response are believed to play a role in influenza infection pathology. Host targeted therapies are promising therapeutics but their effectiveness is limited due to the lack of a systematic characterization of the immune system during infection.


Transcriptional profiling is a promising approach to multi-scale characterization of the intracellular signaling and immune cell population changes that occur during disease. While traditional statistical tests work well for identifying genes and signaling pathways differentially expressed in diseased tissue, they are not well suited for evaluating the effect of immune cell migration on gene expression.

Deconvolution algorithms seek to exploit gene expression to infer cellular demographics. Basic deconvolution algorithms postulate that expression profiles of cell mixtures are linear combinations of the expression profiles of pure cells, and use linear regression tools to enumerate cell proportions. These algorithms have been utilized to de-convolute mixtures of different kinds of tissues or cells [2], subdivisions of immune cells [3], or immune cells at different states (resting or activated) [3]. Most of these algorithms are only tested by artificial data of easily distinguishable mixtures, such as mixtures of lung, liver and brain [2], or mixtures of T cells, monocytes, B lymphoblasts, and B-cell lymphomas [3]. Furthermore, the cell proportions can only be accurately determined if all cell types of the sample are known a priori. Other approaches, such as CTen, use dynamic clustering analysis of time-course transcriptomic data to identify groups of genes sharing common expression patterns that can be associated with specific immune cells [4]. This method shows high accuracy but can only determine relative fold changes in immune cells (as opposed to fraction of cell population or cell counts). Several tools, including CIBERSORT by Newman et al [5] and DCQ by Altboum et al [6], are designed for complex tissues and temporal samples. However the reported estimations by these tools may not be satisfactory, and the error rates of deconvolution algorithms in cell quantification of complex and time course samples are still not understood.

Methods and results

Here, we evaluate the ability of several computational tools to infer immune cell counts using time-course, gene expression data. We constructed a library of microarray data from multiple published datasets of pure immune cells to obtain their transcriptomic signatures. The temporal microarray data of infected lung tissue [7] are used to evaluate predictions of the deconvolution tools. Pearson correlation coefficients are calculated by comparing estimated abundances at different time points to cell count data determined by fluorescence-activated cell sorting. We find that current approaches result in discontinuities and dramatic oscillations of immune cell proportions in infected tissue over time.


Current deconvolution models demonstrate that transcriptomic signatures of pure cells can be used to estimate the composition of cell mixtures. However, predictions of cell populations’ dynamics during the immune response are not continuous and oscillate heavily due to neglecting dependencies between each time point. Our future algorithm will focus on dynamic models of cell infiltration to better predict immune cell demographics from transcriptional data.


[1] Centers for Disease Control and Prevention, 2016. [Online]. Available: https://www.cdc.gov/flu/about/qa/disease.htm. [Accessed 3 2017].

[2] Y. Zhong, Y.-W. Wan, K. Pang, L. M. Chow and Z. Liu, "Digital sorting of complex tissues for cell," BMC Bioinformatics, vol. 14, no. 89, 2013.

[3] A. Abbas, K. Wolslegel, D. Seshasayee, Z. Modrusan and H. Clark, "Deconvolution of Blood Microarray Data Identifies Cellular Activation Patterns in Systemic Lupus Erythematosus," Plos One, 2009.

[4] J. E. Shoemaker, T. J. Lopes, S. Ghosh, Y. Matsuoka, Y. Kawaoka and H. Kitano, "CTen: a web-based platform for identifying enriched cell types from heterogeneous microarray data," BMC Genomics, 2012.

[5] A. M. Newman, C. L. Liu, M. R. Green, A. J. Gentles, W. Feng, Y. Xu and C. D. Hoang, "Robust enumeration of cell subsets from tissue expression profiles," Nature Methods, p. 453–457, 2015.

[6] Z. Altboum, Y. Steuerman, E. David, Z. Barnett-Itzhaki and L. Valadarsky, "Digital cell quantification identifies global immune cell dynamics during influenza infection," Molecular Systems Biology, vol. 10, no. 2, 2014.

[7] J. E. Shoemaker, S. Fukuyama, A. J. Eisfeld, D. Zhao, E. Kawakami and S. Sakabe, "An Ultrasensitive Mechanism Regulates Influenza Virus-Induced Inflammation," PLoS Pathogens, 2015.