PMCID
stringclasses 24
values | Title
stringclasses 24
values | Sentences
stringlengths 2
40.7k
|
|---|---|---|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Mice were sacrificed 6 weeks after virus injection.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Mice were transcardially perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Brain samples were extracted and cryoprotected in 20% sucrose/4% PFA, immersed sequentially in 20% sucrose (in 4% PFA) and 30% sucrose (in 0.1 M phosphate buffer, PB) until sunk, and then transferred to 30% sucrose/PB for more than 24 h. Brain samples were flash-frozen on dry ice and sectioned at 30 μm on a cryostat (Leica, SM2010R).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For dual-color retrograde virus tracing, brain slices were blocked in 10% donkey serum and 0.3% Triton X-100 at 37 °C for 1 hr.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Slices were then incubated with primary antibodies against green fluorescent protein (GFP, 1:500, Nacalai, 04404–84, RRID: AB_10013361) and tdTomato (1:500, OriGene, AB8181-200, RRID: AB_2722750) at room temperature for 2 hr, then 4 °C overnight.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Slices were washed three times using PBS and incubated with Hoechst 33342 (1:1000, Lifetech, H3570), as well as secondary donkey anti-rat Alexa Fluor 488 antibodies (1:800, Invitrogen, A21208) and donkey anti-goat Alexa Fluor 568 antibodies (1:800, Invitrogen, A11057) at room temperature for 1 hr.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Slices were washed three times using PBS and coverslipped.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Stained slices were imaged with a 4 X objective with numerical aperture 0.16 as a map, followed by 1.5 µm increment z stacks with a 10 X objective with numerical aperture 0.4 (FV3000, OLYMPUS).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Composite images were automatically stitched in the X-Y plane using ImageJ/FIJI.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
RNA FISH experiments were performed using RNA-Scope reagents and protocols (ACD Bioscience, CA), following instructions for fixed-frozen tissue.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For experiments using RNA-Scope, immunohistochemistry was performed following RNA-Scope.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Probes of RNA-Scope used in this study include, Mm-Syt6 (449641), Mm-Pou3f1-C2 (436421-C2).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
scRNA-seq data were aligned with the customized mouse reference genome mm10-3.0.0 adding five projection barcodes as separate genes.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Further projection barcode expression was obtained as described in (Projection barcode library preparation and Projection barcode FASTQ alignment).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
scRNA-seq data were demultiplexed using the default parameters of Cellranger software (10x Genomics, v3.0.2).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Obtained filtered transcription count matrix was used for downstream analysis.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For unsorted samples, we used three mice with three GEM wells in one Chromium Single Cell 3' Chip (v3).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Among unsorted samples, sample mouse #1 recovered 8040 cells, 447,984,945 read pairs were aligned, mean reads per cell is 55,719, median genes per cell is 2382; sample mouse #2 recovered 7443 cells, 399,187,134 read pairs were aligned, mean reads per cell is 53,632, median genes per cell is 2379; sample mouse #3 recovered 7243 cells, 410,627,696 read pairs were aligned, mean reads per cell is 56,693, median genes per cell is 2385.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For FAC-sorted samples, we used three mice with one GEM well in one Chromium Single Cell 3' Chip (v3).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
FAC-sorted sample recovered 2075 cells, 410,434,792 read pairs were aligned, mean reads per cell is 197,799, median genes per cell is 6533.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Demultiplexing of projection index barcode was performed using deMULTIplex R package (v1.0.2) (https://github.com/chris-mcginnis-ucsf/MULTI-seq, copy archived at mcGinnis, 2023) with modifications.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Briefly, we have revised the MULTIseq.align function to count the UMI of each projection barcode separately.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We adopted a minimal Hamming distance of 2 for the MULTIseq.align function to improve the matching accuracy between detected and designed barcodes.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Tag parameters in ‘MULTIseq.preProcess’ function were adjusted according to our user-defined position of index barcode length and position.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Based on our primer design, the expected format is: cell barcode in Read 1 (bases 1–16), UMI in Read 1 (bases 17–28), and projection barcode in Read 2 (bases 31–45).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The filtered count matrix was analyzed and processed using Seurat and Scanpy, including data filtering, normalization, highly variable genes selection, scaling, dimension reduction, and clustering (Stuart et al., 2019; Wolf et al., 2018).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
First, scRNA-seq data from three samples of unsorted cells and one sample of sorted EGFP-positive cells were created as Seurat object separately; genes with less than three counts were removed and cells with fewer than 200 genes detected were removed.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Second, four Seurat objects were merged using the ‘merge’ function in Seurat.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Downstream analysis of merged Seurat objects were as follows: (1) Data filtering: cells with a mitochondrial gene ratio of greater than 20% were excluded.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We kept cells for which we detected between 500 and 8000 genes (cells with more than 8000 genes detected were considered potential doublets), and between 1000 and 60,000 counts (cells with more than 60,000 counts detected were considered potential doublets). (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
2) Data normalization: for each cell, counts were log normalized with the ‘NormalizeData’ function in Seurat; ‘scale.factor’ was set to 50,000. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
3) Highly variable gene selection: 2000 highly variable genes were calculated using the ‘FindVariableFeatures’ function in Seurat. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
4) Data scaling: the Seurat object was performed using the ‘ScaleData’ function with default parameters.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The number of counts, number of genes, mitochondrial gene ratio, and sorting condition were regressed out in ‘ScaleData’. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
5) Principal component analysis: highly variable genes were used to calculate principal components in the ‘RunPCA’ function.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A total of 100 principal components (PCs) were obtained and stored in Seurat object for computing neighborhood graphs and uniform manifold approximation and projection (umap) in following section. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
6) Leiden clustering: Seurat object was converted into loom file and imported by Scanpy.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A neighborhood graph of observations was computed by ‘scanpy.pp.neighbors’ function in Scanpy.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Then, leiden algorithm was used to cluster cells by ‘scanpy.tl.leiden’ function in Scanpy. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
7) Cluster merge and trimming: The top 200 DEGs for each cluster were calculated using the ‘scanpy.tl.rank_genes_groups’ function in Scanpy using parameters method=‘wilcoxon’ and n_genes = 200.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Cluster annotation was performed manually based on previously reported markers of PFC all cell types, layer, neuron subtypes, and mouse brain atlas (Bhattacherjee et al., 2019; Sorensen et al., 2015).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Cell clusters with similar marker genes were merged into one cluster.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Complete marker lists for all cell types and all excitatory neuron subtypes calculated using ‘FindAllMarkers’ function in Seurat were provided (see Supplementary files 2 and 3).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Two rounds of clustering were performed.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In the first round, we clustered all cells detected by scRNA-seq to generate major cell type classification, that is excitatory neurons, inhibitory neurons, astrocytes, oligodendrocytes, endothelial cells, and microglia.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Then we use the annotated ‘Excitatory neuron’ cluster to further cluster excitatory neuronal subtypes.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In the 2nd round clustering, we found several clusters expressed a lower number of counts per cell, a lower number of genes per cell, a higher percentage of mitochondria genes, and ribosome protein genes as DEGs, which indicates cell clusters with low cell quality (Ilicic et al., 2016).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We also found several other clusters with a small number of cells expressing typical markers of non-neuron cells, such as microglia (C1qa, C1qb) oligodendrocytes (Olig1, Olig2) and endothelial cells (Flt1, Cldn5), which indicated ‘contamination’ of other cell types mixed in ‘Excitatory neuron’ in the initial clustering results.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We then filtered out those cells from ‘Excitatory neuron’ cluster and redid clustering to generate excitatory neuronal subtypes (see Supplementary file 1).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
To evaluate whether the transcriptional cell types we recovered and annotated correlated with cell types from spatial transcriptomics of PFC or other scRNA-seq datasets of PFC, we used a previously reported comparison analysis method (Bhattacherjee et al., 2023).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Briefly, we integrated our dataset and previously reported datasetes (Bhattacherjee et al., 2023; Bhattacherjee et al., 2019; Lui et al., 2021; Yao et al., 2021) into a harmonized PCA space using the Harmony algorithm (Korsunsky et al., 2019).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We then constructed a K-nearest neighbor (KNN) graph incorporating all cells from the two datasets.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We used the first 30 harmonized principal components as inputs for FindNeighbors function of Seurat to calculate the KNN.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For each cluster of public dataset, we found its 30 nearest neighbor cells and determined the percentages of those cells belonging to each scRNA-seq cluster of our dataset.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
This created a correspondence matrix showing the transcriptional similarity of each public dataset cluster to each cluster of our dataset.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In this matrix, rows represent our scRNA-seq clusters, columns represent public dataset clusters, and the matrix values reflect the degree of similarity between the clusters.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
This process was reciprocally conducted for clusters of our dataset, comparing them to public dataset clusters to form a secondary correspondence matrix.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The mean of these two matrices provided a quantifiable measure of the similarity between cell clusters identified by our annotation and public dataset annotation.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
To determine valid barcoded cells, we first calculated the 95th percentile of the total number of unique molecular identifiers (nUMI) that were mapped with five barcodes, and removed the unusually high numbers of UMIs, which might indicate doublets or PCR-biased amplification.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Next, we used two set of cells as negative control, that is, cells supposed not to contain projection barcodes.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
First set of negative control cells we used is non-neuronal cells classified by coarse clustering based on single-cell transcriptome (Tervo et al., 2016).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Second set of negative control cells we used is ‘EGFP-negative’ cells in FAC-sorted dataset.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Basically, we calculated the total five projection barcodes counts determined by cellranger of FAC-sorted dataset, then we assigned the cells with zero projection barcodes (nUMI of EGFP RNA = 0) counts as ‘EGFP-negative’ cells.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For two set of negative control cells, we searched for the value in the empirical cumulative distribution function (ECDF) that is closest to the 99.9th percentile agains each projection barcode, respectively.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We selected the higher UMI threshold from the two given sets of threshold values.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A cell is determined to be validly barcoded if the number of the barcode UMIs within the cell is larger than the threshold.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For example, the calculated threshold of UMIs for barcode 0 (AI) is 28, which means if a cell contains more than 28 UMIs of barcode 0, then this cell is validly barcoded by AI.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
UMIs threshold for DMS, 101; for MD, 114; for BLA, 35; for LH, 103.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Finally, we dropped UMI counts of determined non-barcoded cells to zero to obtain the index barcode counts matrix used for downstream analysis.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Binary projection patterns were calculated by five projection targets set intersections of corresponding barcoded cells.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Only the top 10 frequent binary and collateral projection patterns were kept for reliable inference.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
DEGs were calculated using the default parameters of the ‘FindMarkers’ function in Seurat, except the MAST algorithm was used to do DE testing.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For the DEG volcano plot, the chosen cut-off for statistical significance was 10 (Figure 4 and Figure 4—figure supplement 1) or 10 (Figure 5 and Figure 5—figure supplement 1) and chosen cut-off for absolute log2 fold-change was 0.5.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Volcano plots were implemented using the EnhancedVolcano R package (v1.4.0).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For the DEG heatmap in Figure 5A, the top 10 DEGs ordered by average log2 fold-change were chosen from each binary cluster.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The heatmap was implemented using the ‘scanpy.pl.heatmap’ function in Scanpy.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Single-neuron projectome data for five PFC target regions (AI, dorsal striatum, BLA, MD, LH) were extracted from Gao et al., 2022.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Projection patterns were quantified by calculating the percentage of each pattern relative to total patterns.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Patterns were categorized by number of targets (1, 2, 3, or ≥3 targets).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
MERGE-seq and fMOST projection pattern percentages were statistically compared within each category using two-sided Wilcoxon tests with Holm correction for multiple comparisons.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Naïve Bayes was applied to perform a machine learning classification task.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We first encoded binary projection labels for each projection target (barcoded and non-barcoded) and five set of models (AI, DMS, BLA, LH and MD) were independently trained.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We explored a parameter range of number of the top highly variable genes (HVGs) (2, 5, 10, 20, 50, 100, 200, 300, 400, 500, 1000, 2000, 5000) to fit the model.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A total of 1000 cells were randomly sampled from 8210 excitatory neurons and top HVGs were selected by default order of results based on ‘FindVariableFeatures’ function of Seurat per trial.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In total, 100 trials were repeated.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
To interpret contribution of important genes for each HVGs-based Naïve Bayes model, data matrix for modeling building was constructed as below: for each projection target, 8210 excitatory neurons × (normalized expression of the top 50 HVGs + binary projection labels), or 8210 excitatory neurons× (normalized expression of 50 random genes +binary projection labels).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Each data matrix was shuffled first and split by training-testing data in a ratio of 0.7.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Machine learning workflow was implanted in pycaret python package (v2.3.4) ‘pycaret.classification’ module.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
First, for each model, we used ‘setup’ function to initialize the training environment and created the transformation pipeline by setting ‘target’ parameter to column name of input data matrix corresponding to binary projection labels.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Then we used ‘create_model’ function to train and evaluate the performance of a given model by setting ‘estimator’ parameter to ‘nb’ and other parameters by default.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
To validate barcode/non-barcode label integrity, we performed 100 iterations of random sampling 1000 cells and swapping barcoded with non-barcoded labels.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Prediction accuracy, AUC, and F1 scores were compared between original models using the top 50 HVGs with true labels versus models with swapped labels.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For each of the 100 trials, 1000 cells were randomly sampled from the 8210 total cells, and barcoded/non-barcoded labels were swapped to the extent possible based on the smaller group.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Models were built for each target using original or swapped labels and the top 50 HVGs.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We implemented kernel explainer of SHAP python package (v0.40.1) to summarize the effects of genes.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
SHAP explainer was created using ‘shap.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
KernelExplainer(model.predict, training data)’ function.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
SHAP values were calculated using ‘explainer.shap_values(testing data)’ function, and plotted by ‘shap.summary_plot()’ function to create a SHAP beeswarm plot by displaying top 20 features.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Training data and testing data for calculating SHAP values were subsampled with 1500 cells.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
No statistical methods were used to predetermine sample size.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.