Page 308 - Contributed Paper Session (CPS) - Volume 4

CPS2233 Sharon Lee
by manual analysis. The CCR was calculated separately for each data and the
results are shown in Table 1, together with the results by HDPGMM (Cron et
al., 2013) and FLAME (Pyne et al., 2009). The later method adopts a cluster
matching step in a post-hoc manner. It can be observed from Table 1 that
Hcyto obtained a higher CCR than HDPGMM and FLAME for most of the 16
data. This is also supported by the average CCR across the batch, where Hcyto
obtained 0.929 compared to 0.796 and 0.538 obtained by HDPGMM and
FLAME, respectively.

4. Discussion and Conclusion
The clustering and alignment of cell populations across multiple data is an
interesting and challenging problem. The proposed Hcyto method adopts a
hierarchical approach to automatically segment and match these clusters, with
implicit models that can directly handle non-normal distributional features.
The methodology is motivated and demonstrated by cytometric data analysis,
but is applicable to other types of data with similar structure. Results from the
real example shows that Hcyto provides improved accuracy compared to other
algorithms that adopt intuitive approaches such as pooling and post-hoc
cluster matching. Future work may look at the scalability of the Hcyto
framework for larger data and extend it for use in downstream analyses such
as identification of discriminatory features, supervised classification of
unlabelled data, and longitudinal modelling of batches.

References
1. Aghaeepour N, Nikolic R, Hoos HH, Brinkman RR. (2011). Rapid cell
population identification in flow cytometry data. Cytometry Part A
79A:6–13.
2. Aghaeepour, N., Finak, G., The FLOWCAP Consortium, The DREAM
Consortium, Hoos, H., Mosmann, T., Gottardo, R., Brinkman, R.R.,
Scheuermann, R.H. (2013). Critical assessment of automated ow
cytometry analysis techniques. Nature Methods 10, 228-238
3. Aghaeepour, N., Chattopadhyay, P.K., Chikina, M., Van Gassen, S., Kurs,
M., Malek, M., McLachlan, G.J., Qui, P., Saeys, Y., Stanton, R., Tong, D.,
Wang, K., Nolan, G., Finak, G., Gottardo, R., Mossman, T., Scheurmann R.,
and Brinkman, R. (2016). Benchmark for evaluation of algorithms for
identification of cellular correlates of clinical outcomes. Cytometry: Part
A 89A, 16-21
4. Azzalini, A., Dalla Valle, A.(1996). The multivariate skew-normal
distribution. Biometrika 83, 715726 Cron A, Gouttefangeas C, Frelinger J,
Lin L, Singh SK, et al. (2013). Hierarchical modeling for rare event
detection and cell subset alignment across flow cytometry samples. PLoS
Computational Biology 9.

297 | I S I W S C 2 0 1 9