Cell-type-specific transcriptomes and the Allen Atlas (II): discussion of the linear model of brain-wide densities of cell types

摘要

The voxelized Allen Atlas of the adult mouse brain (at a resolution of 200microns) has been used in [arXiv:1303.0013] to estimate the region-specificityof 64 cell types whose transcriptional profile in the mouse brain has beenmeasured in microarray experiments. In particular, the model yields estimatesfor the brain-wide density of each of these cell types. We conduct numericalexperiments to estimate the errors in the estimated density profiles. First ofall, we check that a simulated thalamic profile based on 200 well-chosen genescan transfer signal from cerebellar Purkinje cells to the thalamus. Thisinspires us to sub-sample the atlas of genes by repeatedly drawing random setsof 200 genes and refitting the model. This results in a random distribution ofdensity profiles, that can be compared to the predictions of the model. Thisresults in a ranking of cell types by the overlap between the original andsub-sampled density profiles. Cell types with high rank include medium spinyneurons, several samples of cortical pyramidal neurons, hippocampal pyramidalneurons, granule cells and cholinergic neurons from the brain stem. In somecases with lower rank, the average sub-sample can have better contrastproperties than the original model (this is the case for amygdalar neurons anddopaminergic neurons from the ventral midbrain). Finally, we add some noise tothe cell-type-specific transcriptomes by mixing them using a scalar parameterweighing a random matrix. After refitting the model, we observe than a mixingparameter of $5\%$ leads to modifications of density profiles that span thesame interval as the ones resulting from sub-sampling.