Tetsuya J. Kobayashi: Inferring and decoding immunological state from high throughput sequencing of immuno-repertoire

Please note the unusual time at *14:00 hours*

Abstract: Our immune system protects our body from invasions of foreign pathogens. In order to recognize a huge variety of pathogens, a variety of immune cells, each being supposed to recognize different antigens, is prepared and maintained. This diversity of the immune cells is an integral part of the immune system, and decoding the diversity (also known as repertoire) is crucial for defining our immunological state, predicting its disorders, and restoring it by immunotherapy.

Recent advancements of high throughput sequencing (HTS) are giving us greater ability to access the diversity. However, sequence data of immuno-repertoire, consisting of millions of different sequences of T or B cell receptors, is intrinsically high-dimensional and sparse.These properties hamper us from analyzing HTS for extracting biologically relevant information.

In this talk, I introduce a dimensionality reduction based method, in which a low dimensional structure within repertoires is extracted by projecting disimilarity relation among sequences via manifold learning [1]. Difference between repertoires is quantified by using kernel density estimate and Jensen-Shannon divergence (JSD), and major contributing sequences to the difference is detected by focusing on the value of local JSD.

The consistency of our method with previous ones is tested by using a data set of genetically engineered mice, and its applicability is demonstrated by using a clinical data of skin cancer. I would also like to discuss about possible extension of our method.

Reference:
[1] Quantification of inter-sample differences in T cell receptor sequences, Ryo Yokota, Yuki Kaminaga, Tetsuya J. Kobayashi, bioXiv.

biorxiv.org/content/early/2017/04/20/128025