Fig. 4: Classification of cell subtypes based on morphological and connectivity features.

a, A pairwise soma-distance map of CP neurons biclustered on the basis of the spatial adjacency of somas mapped to CCFv3. b, Histograms of the pairwise soma distances for neurons in CP, ACAd6a and LGd regions. c, Matrices of morphology-feature similarity scores (m-similarity) and connectivity-feature similarity scores (c-similarity) of individual CP neurons, rows and columns sorted in the same order as in the clustered d-map in a. Cosine similarity scores are used. d, Connectivity-feature-based clustering of CP neurons into three main subclasses (red, blue and green), in the same convention as in Fig. 2c. e, A 3D visualization of the three CP neuron subclasses in d. f, The 3D soma locations of CP neurons. Color indicates the largest Euclidean distance between axonal terminals and the respective soma. g, LGd neuron d-map and respective m- and c-similarity matrices, rows and columns sorted in the same order. h, ACAd6a neuron d-map and respective m- and c-similarity matrices, rows and columns sorted in the same order. i, Histogram of the difference between corresponding c- and m-similarities for all neurons of the 31 s-types in this study. j, Scatter plot and marginal distributions of corresponding c- and m-similarities for all neurons in the 31 s-types (linear least-squares regression, R² = 0.245; Pearson correlation coefficient 0.495, two-sided ***P < 0.0001, with the precise value being P < 2.2 × 20⁻¹⁶). k, Correlations between soma-distance map and c- or m-similarity for all 31 s-types. Corr, Pearson correlation coefficient. l, Overall correlations between neurons’ soma distances and the respective similarities in connectivity features or morphology features. l_m_corr, correlation between location distances and m-similarities; l_c_corr, correlation between location distances and c-similarities. The center line indicates the median (50th percentile). The box bounds represent the interquartile range (IQR) between the 25th (Q1) and 75th (Q3) percentiles. The whiskers extend to values within 1.5× IQR. Outliers beyond whiskers are shown as individual markers. Minimum and maximum values are within whisker bounds (two-sided paired-samples t-test, sample size: n = 31, **P = 5.30 × 10⁻⁴).