Connectome-seq Code

Overview

The connectome-seq workflow consists of two main parts:

a. Single-cell mRNA Sequencing Analysis

i. Standardized Workflow

1. Dataset integration
2. Clustering and quality control
3. Cell type annotation (automatic and manual)

ii. Output

• Cell type information

---

b. Barcode Matching Analysis

i. Customized Workflow

1. Cell Ranger processing
   Raw sequencing files (.fastq) are transformed into BAM files.

   • mRNA features and Pre/PostRNA features come from separate FASTQ files.
   • UMIs that differ within Hamming distance = 1 are automatically collapsed by Cell Ranger.

2. Custom Python collapsing of Pre/PostRNAs
   Pre/PostRNAs with the same cell barcode (CB) that differ by Hamming distance = 1 are further collapsed.

3. (Configurable) Disambiguation of Pre/PostRNA assignment
   If a Pre/PostRNA is linked to more than two CBs, it is assigned to the CB with the highest Pre/PostRNA UMI count after passing quality checks.

   Current criteria:

   • CB #1 PreRNA > 5 × CB #2 PreRNA
   • CB #1 PostRNA > 2 × CB #2 PostRNA
   • CBs ranked by Pre/PostRNA UMI count

4. Matching between nuclei and synaptosome Pre/PostRNAs
   Matching is performed at a specified Hamming distance threshold.

---

Code Usage

1. BC_collapse: Processing raw barcode through collapsing cell barcodes.
2. cseq_NucUMIClean: Collapsing nucleus barcode information based on UMI.
3. cseq_synaptosome: Preprocessing synaptosomal barcode, applying filters to extract cleaned data.
4. cseq_HarmonyIntegration: Integrate different connectome-seq datasets to eliminate batch effect.
5. cseq_Annotation: Annotate the integrated connectome-seq dataset with Allen Institute mouse brain cell type atlas.
6. cseq_Cluster: Cluster the integrated dataset and annotate each cluster from marker gene expression.
7. cseq_Match: Conduct barcode matching based on synapse barcode and cell barcode.
8. cseq_PurkinjeAnalysis: Analyze possible Pons–Purkinje cell connection.
9. cseq_Rfunction: Wrapper for R functions used in cseq_PurkinjeAnalysis.
10. cseq_virusLib: Analyze virus library sequencing results, extracting barcode information.
11. cseq_preseq: Predict virus library barcode diversity.

---

Barcode Processing

Demonstrating with cseq25 dataset.

Step 1. cellranger count

Run one-line commands to quantify features from raw sequencing data.

cellranger count   --id=cseq25-PN-new   --transcriptome=/home/boxuan/data/hdd2/mouse_2022   --libraries=library_PN.csv   --feature-ref=feature_ref_2.csv   --force-cells=21000   --localmem 128   --localcores=32   --check-library-compatibility=false

cellranger count   --id=cseq25-CN2   --transcriptome=/home/boxuan/data/hdd2/mouse_cseq6   --libraries=library_CN.csv   --feature-ref=feature_ref_2.csv   --force-cells=4000   --localmem 128   --localcores=32

cellranger count   --id=cseq25-CS-pre2   --transcriptome=/home/boxuan/data/hdd2/mouse_cseq6   --libraries=library_CSpre2.csv   --feature-ref=feature_ref_pre.csv   --expect-cells=8000   --localmem 128   --localcores=32

cellranger count   --id=cseq25-CS-post2   --transcriptome=/home/boxuan/data/hdd2/mouse_cseq6   --libraries=library_CSpost2.csv   --feature-ref=feature_ref_post.csv   --expect-cells=8000   --localmem 128   --localcores=32

Instructions

1. Change id, libraries, and feature-ref as needed.
2. Use --force-cells for nucleus libraries.
3. Use --expect-cells for synaptosome libraries.

---

Step 2. cellranger subset-bam_linux

gunzip barcodes.tsv.gz

~/cellranger/subset-bam_linux   --bam ../../cseq1-PN2/outs/possorted_genome_bam.bam   --cell-barcodes barcodes.tsv   --cores 32   --out-bam pons.pre.barcodes.bam

Instructions

1. Navigate to the outs/ directory of each cellranger count result.
2. Unzip barcode files:

   gunzip barcodes.tsv.gz

3. Create a new folder for each sample.
4. Copy:
   • Filtered barcodes.tsv from nucleus samples
   • Unfiltered (raw) barcodes.tsv from synaptosome samples

---

Step 3. connectome.py

../connectome.py pons.pre.barcodes.bam pons.pre.barcodes.hamm1.txt 1

Instructions

1. Extract reads and collapse barcodes at a given Hamming distance.

---

Step 4.1. cseq_synaptosome.Rmd

Instructions

1. Synaptosome barcode information (TXT format) is processed with cseq_synaptosome.Rmd, where extensive quality control steps are taken. Output:
   • synp.post.clean.csv
   • synp.pre.clean.csv
     These data serve as input for cseq_Match.Rmd.

Step 4.2. cseq_NucUMIClean.ipynb

Instructions

1. Cell barcode information (e.g., cere.post.BC.hamm1.txt, pons.pre.BC.hamm1.txt) are UMI collapsed with cseq_NucUMIClean.ipynb, producing:
   • CN.umi.csv
   • PN.umi.csv
     These serve as input for cseq_Match.Rmd, where cell barcodes undergo disambiguous assignment and cleaned barcode information is generated.

---

Step 5. cseq_Match.Rmd

Instructions

1. Follow code annotation to run this RMarkdown.

---

Transcriptome Processing

1. Raw single-nucleus sequencing data were processed with CellRanger to extract feature_bc_matrix files:
   • barcodes.tsv.gz
   • features.tsv.gz
   • matrix.mtx.gz
     These are used as input for cseq_HarmonyIntegration.Rmd, where Seurat objects are created and biological replicates integrated with Harmony. Output Seurat objects:
   • cn.af
   • pn.af
2. Integrated data go through cell type prediction by MapMyCell in cseq_Annotation.Rmd, generating:
   • scdata.pn
   • scdata.cn
     Further QC, dimension reduction, and annotation are performed in cseq_Cluster.Rmd, producing:
   • celltype.pons.new
   • celltype.cere.new
   • pn_barcodes
   • cn_barcodes
     These provide cell type information for different CBs and corresponding barcodes, serving as input for cseq_Match.Rmd.

---

Connectome Processing

All outputs above serve as input for cseq_Match.Rmd, which generates the final connectome. Outputs include:

• PN.neuron.umiClean.csv
• CN.neuron.umiClean.csv
• synp.pre.clean.csv
• synp.post.clean.csv
• connectome.match.count.csv

---

Purkinje Cell Analysis

Purkinje cells are extracted from annotated snRNA-seq data and processed with cseq_PurkinjeAnalysis.Rmd to analyze potential connectivity-related gene markers.

---

Virus Library Analysis

1. Raw sequencing data of Barcode AAV genomic DNA are processed with cseq_VirusLib.ipynb. Barcode sequence and count information are extracted, and distribution calculated.
2. cseq_preseq.sh is applied to project barcode sequence diversity at future sequencing depth.

---