Labpib:PCA

From OpenWetWare

Jump to: navigation, search

Click here to see the *official* version of this page Clique aqui para ver a versão *oficial* desta página

Image:Usp_logo_header_LabPIB.jpg

Home|Apresentação    Research|Pesquisa    Teaching|Ensino    Outreach|Extensão    Members|Membros   

Publications|Publicações    Contact|Contato    Join the Team|Junte-se ao Grupo    Lab Policy|Regras do Lab   

PCA analysis of single cell type transcriptomes

The idea is to project data points (transcriptomes) from a high-dimensional space (transcription levels of several genes) into a lower-dimensional (3d to be easy to see) space. This projection is not arbitrary but rather unique... it is the Principal Component projection. However, instead of using all the samples (transcriptomes) available to perform a traditional PCA analysis, the rotation+projection involved in a PCA analysis is performed only using some basic types of prostate cells: Luminal, Basal, Endothelial and Stromal. These single-type cells were obtained by cell-sorting and their transcriptomes measured, hoping to de-convolute a specific signal (expression levels by Affy arrays) for each cell type that are usually confounded if one measures bulk tissues. Using only these basic cells we hope to create a special projection space as if one was looking from the point-of-view of what distinguishes (the principal components) each basic cell-type that make a prostate tissue. Given this sub-space (just the 3 first dimensions after PCA) one can now rotate+project all other transcriptome samples (stem-cells, cancer cells, bulk tissues, cell cultures, etc) into it and "see" the differences and similarities among them from the perspective of basic cell-types "modes".

The R-language script that make this analysis possible is available, along with some illustrative example data, at this link: script-PCA.zip

Personal tools