Journal: eLife

Article Title: Cell type composition and circuit organization of clonally related excitatory neurons in the juvenile mouse neocortex

doi: 10.7554/eLife.52951

Figure Lengend Snippet: ( A ) Overview of experimental approach using Patch-seq. ( B ) Example tdTomato-positive translaminar clone spanning cortical layers 2–6 in an acute cortical slice used for Patch-seq experiments. Overlay of bright field and fluorescence image was performed in Adobe Photoshop. Scale bar: 100 μm. ( C and D ) Box plots showing library size ( C ) and number of genes detected ( D ) for all cells passing quality control criteria (n = 206). ( E ) Density plot of the percent of variance in normalized log-expression values explained by different experimental factors. Each curve corresponds to the variance in gene expression across all genes (n = 12,841 genes) that can be explained by a single variable, with right-shifted curves reflecting variables that explain a higher fraction of the variance. ( F ) T-distributed stochastic neighbor embedding (t-SNE) plots using the top highly variable and correlated genes across all cells (n = 91 genes; n = 87, 22, 84, and 13 cells in layers 2/3, 4, 5, and 6, respectively), colored by layer position. ( G ) Performance of a generalized linear model (GLM) trained to predict region from gene expression data of L2/3 neurons (n = 12,841 genes and 85 cells) with model performance (black dot) compared to the chance-level performance estimated using shuffled data (gray, mean and 95% coverage interval; one-tailed p -value computed from shuffled data, shuffling region). ( H ) Performance of a GLM trained to predict region from gene expression data of L5 neurons (n = 12,841 genes and 77 cells) as described in ( G ). See also and and . Figure 2—source data 1. Gene expression data, related to . Normalized counts, normalized log counts, and metadata for all Patch-seq neurons included in our analysis.

Article Snippet: We fit a binomial GLM (using the glmfit() function in Matlab) containing the relevant linear terms and all possible pairwise interactions: g ( P ) = β 0 + β L ⋅ L + β C ⋅ C + β D ⋅ D + β R ⋅ R + β L C ⋅ L ⋅ C + β L D ⋅ L ⋅ D + β L R ⋅ L ⋅ R + β C D ⋅ C ⋅ D + β C R ⋅ C ⋅ R + β D R ⋅ D ⋅ R where β 0 is a constant term, L is a binary variable representing the lineage relationship (1 for related and 0 for unrelated), C is a binary variable representing the connection type (1 for vertical and 0 for lateral), D is the Euclidean distance between the cells in microns, R is a numeric variable representing the rostrocaudal position of the clone with integer values from 1 (most rostral) to 5 (most caudal), and β i are the corresponding coefficients.

Techniques: Fluorescence, Control, Expressing, Gene Expression, One-tailed Test

Journal: eLife

Article Title: Cell type composition and circuit organization of clonally related excitatory neurons in the juvenile mouse neocortex

doi: 10.7554/eLife.52951

Figure Lengend Snippet: Generalized linear model of connectivity. Connectivity was modeled as a binomial response variable with the following predictors: lineage relationship (1 for related, 0 for unrelated), connection type (1 for vertical, 0 for lateral), Euclidean distance between the cells in microns, and rostrocaudal position (a numeric factor from 1 to 5; see Materials and methods). ‘×’ denotes an interaction between two linear terms. Overall χ 2 = 33.5 compared to constant model, p=2.26 × 10 −4 , 1988 error degrees of freedom. The four terms with small p -values are: connection class (connection probability P is lower for unrelated vertical connections, compared to unrelated lateral), Euclidean distance (P decreases with increasing distance for unrelated lateral connections), lineage × connection type ( P is higher for related vertical pairs), and connection type × Euclidean distance (the effect of Euclidean distance on P depends on the type of connection tested).

Article Snippet: We fit a binomial GLM (using the glmfit() function in Matlab) containing the relevant linear terms and all possible pairwise interactions: g ( P ) = β 0 + β L ⋅ L + β C ⋅ C + β D ⋅ D + β R ⋅ R + β L C ⋅ L ⋅ C + β L D ⋅ L ⋅ D + β L R ⋅ L ⋅ R + β C D ⋅ C ⋅ D + β C R ⋅ C ⋅ R + β D R ⋅ D ⋅ R where β 0 is a constant term, L is a binary variable representing the lineage relationship (1 for related and 0 for unrelated), C is a binary variable representing the connection type (1 for vertical and 0 for lateral), D is the Euclidean distance between the cells in microns, R is a numeric variable representing the rostrocaudal position of the clone with integer values from 1 (most rostral) to 5 (most caudal), and β i are the corresponding coefficients.

Techniques:

Journal: PLoS ONE

Article Title: Deep Brain Stimulation Imposes Complex Informational Lesions

doi: 10.1371/journal.pone.0074462

Figure Lengend Snippet: A: Microelectrode recordings were performed in regions of the globus pallidus and thalamus with spike activity that was responsive to passive joint movement. B: Results of experimenter-blinded muscle rigidity scoring for both monkeys at three DBS settings. C and D: Co-registration of pre-operative MRI and post-electrode implantation CT showing DBS electrode location for monkey R (C) and K (D). E and F: Localization of recorded cells obtained from stereotactic navigation software and overlaid on corresponding atlas plates for monkey R (top) and K (bottom) for both the pallidum (E) and the thalamus (F). G: A generalized linear model (GLM) accounting for position, velocity, and acceleration of the joint movement was applied to determine the correlation between kinematics of the joint movement (top row) and spike activity (2 nd row: spike raster, 3 rd row: corresponding rate histogram). Bottom row shows the GLM prediction of firing rate.

Article Snippet: A generalized linear model (GLM) fit function in Matlab (Mathworks, Natick MA) was applied, with Δt=1 ms and covariates of position, velocity, and acceleration in the plane of the tracked limb’s movement ( ).

Techniques: Activity Assay, Software