Fig 1.
Creation of top projections from coronal images of gene expression.
(A) Image of a coronal section from the Allen Mouse Brain Atlas. Gene: Rorb. ISH intensity is normalized to a range of 0 to 1. Color scale shown in panel C. (B) Binary cortex mask with value = 1 for cortical pixels and value = 0 for subcortical pixels. (C) Product of the images in A and B, resulting in ISH intensity values in cortical pixels and zero’s in subcortical locations. (D) Schematic illustration of the projection process used to generate top projections. (E) Top projection for Rorb. Dashed line indicates location of section in panel A. (F) Cortical boundaries from the Allen Mouse Brain Reference Atlas, overlaid onto the gene expression top view of panel E. (G) Schematic illustration of the projection process used to generate flat map top projections. (H) Flat map projection for Rorb. Dashed line indicates location of section in panel A. (I) Cortical boundaries from the Allen Mouse Brain Reference Atlas, overlaid onto the gene expression top view of panel H.
Fig 2.
Random forest algorithm: Method and example results.
(A) A coronal image (left) and the top projection (right) for gene Nvl. Note the missing data (black pixels). (B) Top view for gene Adra1d. Note the pronounced variation in density along the a-p axis. (C) Binary mask for primary somatosensory cortex barrel field (SSp-bfd). Light gray inside SSp-bfd; darker grey marks pixels in surrounding region. White lines: boundaries of cortical areas in Allen Mouse Brain Reference Atlas. (D) Schematic illustration of arrays input into Random Forest algorithm. Columns correspond to gene, rows to pixels in the top projection data set. Each value is an ISH luminance value. Classification of pixel is taken from the reference mask (panel C). (E) Confusion matrix output from Random Forest algorithm for SSp-bfd. 0 indicates point outside SSp-bfd, 1 indicates point inside SSp-bfd. (F) Gene importance histogram. Importance values approximate a logarithmic distribution. (G) Classification accuracy for all cortical areas (top and flat-map projections) shows no increase in accuracy with increasing area size. (H) Examples of genes that mark SSp-bfd, with overlaid Allen Mouse Brain Reference Atlas borders. Nov rank 1, importance 0.022. Hlf rank 10, importance value 0.0081. Stap2 rank 100, importance 0.0018.
Table 1.
List of potential genetic markers.
Potential cortical boundary genetic markers, listed by cortical region, as identified by Random Forest variable importance classifier. All explored regions listed. Regions with no listed genes displayed no clear potential genetic marker. Each gene was identified independently.
Fig 3.
Examples of markers for cortical areas.
(A) R-Spondin 1 labels primary somatosensory cortex barrel field. (B) Wnt Family Member 7B labels primary motor cortex. (C) Retinoid-Related Orphan Receptor, Beta labels primary sensory areas. (D) Transmembrane Protein 215 labels dorsal retrosplenial cortex. (E) Cadherin 24 labels primary auditory cortex. (F) Neuropeptide S Receptor 1 labels retrosplenial cortex. (G) Serpin Family F Member 1 labels frontal pole. (H) Leukemia Inhibitory Factor Receptor labels temporal association cortex. Panels A-F provide examples of genes identified in top projections, panels G and H of genes identified in flat maps. Cortical area of interest is marked with cyan border.
Fig 4.
Single cell RNA-sequencing expression plot.
Expression of top-projection identified potential areal marker genes in mouse cortical cells grouped into subtypes of three major cell classes: inhibitory neurons, excitatory neurons, and non-neuronal cells. Expression data was measured by RNA-sequencing of single cells isolated from Primary Visual Cortex (VISp) and Anterior Lateral Motor Cortex (ALM). Expression is taken from Smart-seq version 4, as log-transformed average counts per million (CPM) of intronic and exonic reads. Color corresponds to expression value, with warmer colors indicating high expression and cooler colors indicating low expression. Max value indicates the maximum expression per gene, measured in average CPM per cluster. CPM = counts per million reads.