[placeholder for image]

Hindbrain Explorer

Pfister Lab Kutscher Lab Kaessmann Lab More apps

This app accompanies the paper A multi-omics atlas of human hindbrain development.

This interactive app allows users to explore the data presented in the study including:

  • gene expression,
  • chromatin accessibility,
  • transcription factor activities,
  • and gene regulatory networks
In addition, users can apply our deep learning model deepHB to predict the activity of sequences in hindbrain cell types and interpret the model's decisions, by showing base-resolution contribution scores. Finally, users can upload their own single-cell RNA-seq data of human hindbrain data and let our classifier predict cell type labels for their cells.

Here a user can explore the hindbrain scRNA-seq atlas by a) plotting the UMAP embedding of the integrated atlas (All Classes) or select one of the 16 classes, with cells colored as per various metadata features, b) plotting the expression of the main marker genes in the annotated clusters in the integrated atlas or selected class, c) plotting the proportional contribution of developmental stages and sex to cells comprising each of the clusters.

Here a user can explore the expression of selected gene(s) in all the annotated clusters (All Classes) or chose a particular class to view the expression in either cluster or subcluster level within that class.

Here a user can explore the 446,510 peaks identified in the integrated hindbrain snATAC-seq data in their genomic context and check their chromatin accessibility across different cell clusters. The table also lists if a particular peak was identified as a marker peak for a class and/or for cluster(s). Please note that marker peaks for clusters within a class were identified using within class comparison. To get started, select a cis-regulatory element (CRE) from the table below. 


                    

                    

                      
                      

                        
                        
                        
                        
                      

                    

                  

                  

                    
                  

                

                

                  
                

              

            

          

          

            

              Here a user can explore the top transcription factor gene regulatory network (TF-GRN)
              programs identified in each cell class at the cluster or subcluster level. The top plot
              depicts the enrichment of each TF-GRN program across clusters or subclusters. The bottom
              plot shows the expression of the corresponding TFs for the TF-GRN programs across clusters
              or subclusters (Top TFs tab). A user can also select custom TFs to see their expression
              across clusters or subclusters (Custom TFs tab) in a selected class.
            

            

              

                
              

              

                

                

                  
                  

                    

                      

                    

                    

                      

                        

                          

                            
                            

                              
                              
                            

                          

                        

                      

                      

                    

                  

                

              

            

          

          

            

              Here a user can explore context specific activity of TF across multiple classes. Positive
              correlation (top plot) shows conserved activity as the gene-sets associated with the TF in
              shown classes have correlated expression. Negative or zero correlation suggest context
              specific activity as GRNs associated with the TF doesn’t have correlated expression. The
              AUC plots (middle and bottom) depicts if the activity of a TF-GRN is statistically enriched
              in a particular class.
            

            

              

                
              

              

                

              

            

          

          

            

              Here a user can utilize our deep learning-based model, DeepHB, to predict the
              meta-regulatory program (rMP) class of a selected region of interest (ROI). A user can
              provide either genomic coordinates or DNA sequence for the ROI. Please not that the
              selected ROI has to be 500 bp in length which is the length of the sequences used for
              training the model.
            

            

              

                
              

              

                

                
Contribution scores

                

                
Top 5 sequence motifs per MP

                

                  

                    

                  

                  

                    

                  

                

                

                

                  

                    

                  

                  

                    

                  

                

                

                

                  

                    

                  

                  

                    

                  

                

                

                

                  

                    

                  

                  

                    

                  

                

                

                

                  

                    

                  

                  

                    

                  

                

              

            

          

          

            

              Here a user can upload their own single-cell gene expression data from any human data-set
              and predict cluster and class labels for individual cells. For each cell, the best matching
              label is identified and then using a cut-off of 0.4 prediction score, labels with lower
              confidence are assigned “ND” identity. The plot shows distribution of class and cluster
              composition of the given sample. For cluster, only top 10 most frequent cluster identities
              are shown.
            

            

              

                
              

              

                

                

              

            

          

          

            

              
Information & Data

              
Reference

              

                A multi-omics atlas of human hindbrain development.
                

                Piyush Joshi, Mari Sepp, Ioannis Sarropoulos, Nils Trost, Konstantin Okonechnikov,
                Tetsuya Yamada, Céline Schneider, Julia Schmidt, Ashwyn A. Perera, Andrea Wittmann,
                Mirjam Blattner-Johnson, Barbara Jones, Cornelis M. van Tilburg, Olaf Witt,
                Steven Lisgo, Miklós Palkovits, David T.W. Jones, Supat Thongjuea, Henrik Kaessmann,
                Stefan M. Pfister, Lena M. Kutscher
              

              
Abstract

              

                The human hindbrain controls essential motor and autonomic functions and is the site
                of neurodevelopmental diseases. Yet, its cellular diversity, developmental
                trajectories and underlying regulatory logic remain poorly understood. We present a
                comprehensive multi-omics atlas of human hindbrain development spanning embryonic to
                adult stages, encompassing 594,817 transcriptomic and 422,568 chromatin-accessibility
                single-nucleus profiles. This dataset resolved the cellular architecture of hindbrain
                cellular lineages, and delineated coordinated gene expression, cis-regulatory
                programs and regulatory grammar guiding their developmental trajectories. By
                integrating multi-omics data, we discovered context-specific roles of transcription
                factors across cell types and deciphered the role of HOX genes in driving divergent
                cellular identity in related lineages. We further leveraged the atlas to
                contextualize pediatric gliomas to decode how subtle yet coordinated shifts in gene
                expression context can define oncogenic transformation. Together, this atlas provides
                a foundational resource for hindbrain biology and establishes a gene-regulatory
                framework linking development and disease.