Summary of statistics

Regulatory regions:

• Consensus peaks across ages (i.e. includes brain + VNC): 95,921 (500bp) and 207,325 (150bp) peaks, covering aprox. 39% of the genome. (Note that these peaks use a very lenient peak calling to include potential enhancers from small cell populations).

• Peaks for adult cell types (with more stringent threshold): 60,210 disjoint peaks of an average width of 455 bp, covering 19% of the genome.

• Cell type specific accessibility: - Coming soon

  See also: Extended Data Fig. 3 in the paper, for the number of DARs per cell type.

• Peaks across development: - Coming soon

  See also: Extended Data Fig. 3 in the paper.

• Adult cell types in late pupa: - Coming soon

Overlap with Janelia regions:

• 220 Janelia lines do not overlap with any of the 95k peaks in our dataset (adult+devel), 685 overlap with only one peak, and 2551 with more than one.
  

See also: Tab “Janelia” in “Enhancer Architecture”.

Regulatory domains (i.e. search space for enhancer-gene links):

• Genomic regulatory blocks (GRB): Median GRB size: 127.6kbp; Number of genes in GRBs: 1438 (15% of the 9513 genes with links); Median percentage of links per gene in the same GRB: 100% (average: 80%)

• Topologically associated domains (TAD): Median TAD size: 13.1k (IQR: 7355-28095), Number of genes with their TSS in TADs: 8620 (90% of the 9513 genes with links); Number of genes with their biggest transcript within one TAD: 6410 (67% of 9513); Median percentage of links within TADs (regulatory region and TSS in the same TAD): 14% (average: 26%)

• BEAF-32-defined domains: Median BEAF-32 domain size: 57.7k (median distance between two of the 2878 BEAF-32 peaks with motif: 5296bp, IQR: 23089–105978); 88% of genes are between two BEAF-32 Chip-seq peaks within 200kbp (46% within 50kbp; 86% of genes are within 50kb of a BEAF-32).

• Median distance from GRB to closest BEAF-32 peak: 14.9k upstream, 21.3k downstream. 58% of the GRBs do not include any BEAF-32 peak (and only 19% include more than 1 peak).

See also: Extended Data Fig. 14 in the paper.

Region-gene links:

Within **100kbp* of the gene (from 50kbp upstream to 50kbp downstream, including introns; ignoring other types of boundaries/domains)*

Median number of links per gene (disjoint regions): 6 positive (IQR: 3-11), 3 negative (IQR: 1-5). Most regions are linked to max. 1 gene.

• Of these, a median of 56% (6 links) are within its BEAF-32 domain; with a median distance between the center of the region and the TSS is 23kbp (IQR: 5-33kbp).

• About a 14% of the linked regions are promoters (TSS-500bp), this percentage remains similar between DARs, cistromes and eGRNs:

  • 13% of regions in the eGRN are in a promoter* of the gene they regulate, 43% within the longest transcript, and the rest distal (up to 50kbp from the gene). (8% if only the most upstream promoter per gene).
  • 14% of regions in cistromes are in the promoter of a gene (not necessarily linked), and 67% within a transcript.
  • 14% of DARs are in the promoter of a gene, and 69% within a transcript.

  (*) The categories are not exclusive. i.e., if a TSS is within the transcript of the longest isoform, it is counted twice. Also, for cistromes and DARs a region can be in the promoter of a gene, and transcript of another…

Cistromes:

• 116 TFs have “chromatin-opening” cistromes (positive correlation between TF expression and motif enrichment) with at least 10 target regions. In addition, 63 TFs are also expressed in cell types which present enrichment of their binding motif, but not necessarily with strong correlation (i.e. the TF activity might be regulated at post-transcriptional levels). Of the 116 TFs, 60-80 are neuron-type specific, 10 TFs show pan-neuronal activity across central brain and optic lobe, and 22 are mostly glial.
• 131 TFs present negative correlation between the TF expression and motif enrichment (i.e. these include candidate represors for closing the chromatin).
• 51% of cistrome regions are linked to a target gene, while 19% are retained in the eGRNs. Most of the regions in cistromes are located within 50kbp of expressed or marker genes of the corresponding cell type (76-95%), indicating that our linking method is likely conservative –rather than non-linked regions being orphan enhancers (e.g., without target gene)–.

See also: Extended Data Fig. 8 in the paper, for an overview of the cistromes.

Gene-regulatory networks (eGRNs) - Stats for activator TFs:

• 87 TFs in eGRNs (80 with at least 10 target genes in a cell type), with 4972 enhancers linked to 2023 genes (13% of these linked regions are promoters, 43% intronic, and the remaining more distal). These cover 17% of the adult DARs, and 39% of the marker genes.

• Regulators per gene (across all cell types): an average of 5 TFs and 3 regions regulate each gene (regions IQR: 1-4, range: 1-55; 62% of the genes are regulated by several regions).

• Targets per TF: 61 genes (IQR: 31-132), 84 regions (IQR: 35-204). 31/80 TFs are auto-regulatory.

• Cell type specific eGRNs: Each cell type has a median of 5 TFs (range: 1 to 15) collectively regulating 65 target genes (range: 21-788) through 106 enhancers (range: 19-1171).

• Regulators per gene (within a cell type): most genes are regulated by 1-2 regions (40% of the genes are regulated by several regions). 63% of the genes are regulated by several TFs within the cell type, 93% of them have an enhancer with several TFs.

List of resources

This page contains the list of files available for download

scATAC-seq raw data (fastq files) can be downloaded from GEO (GSE163697), which also includes bigwig, and cisTopic objects.

Pre-processed data per run (output from Cell Ranger), including:

• CellRanger report (web_summary.html)

• Fragments files (fragments.tsv.gz)

• Cell QC stats (singlecell.csv; barcodes with value in the column “cell_id” are considered cells)

Accessibility and expression matrices:

# R code:
library(SCopeLoomR)
loom <- open_loom("AdultP72.loom")
atacMat <- get_dgem(loom)

cto <- readRDS("AdultP72_cisTopic.Rds")
library(cisTopic)
# use cisTopic normally, e.g.:
plot(cto@dr$cell[[1]])
predMat <- predictiveDistribution(cto)

• scRNA-seq count matrices and analisys results:

  • Adult: Davie_Janssens_Koldere_et_al_2018_AdultBrain.loom (118687 cells, 11652 genes; 495 MB)
  • Larva: Ravenscroft_et_al_2019_LarvalBrain.loom (5056 cells, 9853 genes; 20 MB)

• scATAC-seq

  • Count matrix (cntMat_scATAC_240919c_129078r.mtx.gz) (240919 cells, 129078 regions; 2255 MB);
  • Count matrix cells (scATACcells_cntMatRows.txt.gz)
  • Count matrix regions (scATACregions_cntMatCols.txt.gz)
  • Cell info (scATAC_cellData_240919c.tsv.gz)
  • Adult + 72APF: cisTopic probabilities (AdultP72.loom) (31k cells, 57,190 regions; 838 MB); cisTopicObject (AdultP72_cisTopic.Rds) (60,624 cells, 128,927 regions; 127 MB);
  • Adult + 72APF + 48 APF: cisTopic probabilities (AdultP72P48.loom) (88,331 cells, 37,934 regions; 1.8 GB)
  • Larva to 12h APF: cisTopic probabilities (L3P12.loom) (XX cells, XX regions; 2 GB); cisTopicObject (L3P12_cisTopic.Rds) (135,275 cells, 129,027 regions; 233 MB)
  • All time points: cisTopic probabilities (All_timepoints.loom) (150,000 cells, 32,305 ctx regions; 2.5 GB); cisTopicObject (AllTimepoints_cisTopic.Rds) (240,919 cells, 129,078 regions; 476 MB);

Cell-type aggregated files, peaks, and enhancer-gene links (also available for exploration in UCSC Genome browser):

• Cell type aggregates:

  • Adult cells: bigwig/AdultCellTypes/ (bigWig/.bw)
  • Early development (Larva - 12APF): bigwig/EarlyDevelCellTypes/(bigWig/.bw)

• Links enhancer to gene:

  • Positive (predefined regulatory regions): region2geneLinks_pos_ctx.bb
  • Positive (peaks): region2geneLinks_pos_peaks.bb
  • Negative (predefined regulatory regions): region2geneLinks_nonPos_ctx.bb
  • Negative (peaks): region2geneLinks_nonPos_peaks.bb

• Peaks:

  • Adult whole-brain aggregate: BrainPeaks_ResizedToMax500.bed (Regions used for the lvl2 analyses and deep learning)
  • Adult cell types: peaks/AdultCellTypes/ (.bed)
  • Early development cell types(Larva - 12APF): peaks/EarlyDevelCellTypes/ (.bed)

• Other regions:

  • Regulatory regions: ctxRegions.bed (used in cisTopic analyses; source: i-cisTarget)
  • Enhancers tested: enhancersTested.bed
  • BEAF-32 ChIP-seq peaks with motif: BEAF32chip_peaksWmotif.bed

Regulatory networks (cistromes and eGRNs):

• Adult cell types: cistromes_eGRNs_Adult.feather

• Development (only cistromes): 0h, 12h, 24h

Other analysis results - data matrices available in the networks section:

• DARs (Differentially accessible regions for each cell type)

• Motif enrichment

• TFs per cell type

• Regions per motif

• RNA markers

• List of cell types and colors used in the plots: cellTypeInfo.tsv

Other analysis results - Deep learning:

• Overview of all TF patterns accross cell types(TFmodisco, .pdf)

• DeepFlyBrain (DL models, available in Kipoi)

Help & Tutorials

A few examples of queries that can be carried on this website:

-Under construction-

Video abstract

The paper is currently under review. Once we have a final version, we will include a video abstract.

Meanwhile, you can see the pre-print in BioRxiv, and a temporary video presentation here:

Note: the data in the website has been updated during the revision. The networks in the pre-print and presentation are now outdated.

About us

Aerts lab @ VIB - KU Leuven

Contact us:

• Stein Aerts

• Sara Aibar

Bug reports