Single cell analysis¶

On this secction we describe how to perform analysis of single-cell RNA-seq data to quantify microexons across populations of cells and alternative splicing events across them. Since single cell experiments usually provides a shallow sequencing depth for each cell, we have developed a pseudo-pooling strategy to assess differential inclusion of microexons and other types of alternative splicing events across defined groups of cells (normally corresponding to cell-types previously defined by gene expresion profiles). We named this single cell analysis module snakepool an on this secction we describe how to use it.

Note

Before using this module you must follow the same installation instructions decrived in Differential inclusion analysis secction.

Configuration¶

To run this secction the following parameters needs to be incorporated at config.yaml.

Single_Cell : T
cluster_metadata : /lustre/scratch117/cellgen/team218/gp7/Micro-exons/Runs/Paper/MicroExonator/Whippet/Tasic_clustering.txt
cluster_name : broad_type
file_basename : Run_s
cdf_t : 0.8
min_p_mean : 0.9
min_delta : 0.1
min_rep : 25
run_metadata : /lustre/scratch117/cellgen/team218/gp7/Micro-exons/Runs/Paper/MicroExonator/Whippet/Tasic_run.txt

Single_Cell correspond to an optiona parameter that needs to be set as T in order to run snakepool.
cluster_metadata must indicate the path of a tabular separated file that contain at least two colums to indicate the cluster and file base names. This file must have the first row as header.
cluster_name indicate the name of the column, inside cluster_metadata, which has cluster name information.
file_baseame indicate the name of the column, inside cluster_metadata, which has the sample names. These needs to match with sample names defined on the input files (See Setup)
cdf_t parameter set a theshold to run a Cumulative distribution function over the resultant Probability values obtained for each node across comutational replicates, asuming these fit a beta distribution. Which in practical terms can be considered as a user-defined threshold (between 0.5 and 1) to calculate a p-value assosiated to node’s probability of differential inclusion across computational replicates.
min_p_mean corresponds to a threshold of mean probability values across computational samples to define a node as differentially included across the comparing cell-types.
min_p_delta corresponds to a threshold of mean delta PSI values across computational samples to define a node as differentially included across the comparing cell-types.
min_rep minimun set of computational replicates that can be considered to define a node as differentially included across the comparing cell-types. This parameter is relevant because when nodes have limmited read coverage across cells, only some few computational replictates could enable quantitative alternative splicing analyses, leading to unreliable assesment of its alternative inclusion. The number of compuational repeats is defined for each sample inside run_metadata. We recomend to set this value to at least the half of the compuational replicates that are scheduled to run by the user.
run_metadata indicates the path of a tabulat separated file which contain information about user-defined comparisons across cell-types. Additional information about this file can be found bellow.

Warning

If Single_Cell is set to T all the other parameters listed above will become compulsory. Thus, you should only activate this module if you have the required parameters on config.yaml.

run_metadata¶

The file indicated by run_metadata must be a tabular separated file containing the following columns:

**run_metadata.tsv**¶
Column	Description
Compare_ID	User-defined name for scheduled comparions
A.cluster_names	Comma-separated list of cell-types to be concider as part of sample group A
A.number_of_pools	Number of pseudo-bulk to be generated for sample group A
B.cluster_names	Comma-separated list of cell-types to be concider as part of sample group B
B.number_of_pools	Number of pseudo-bulk to be generated for sample group B
Repeat	Node type. For more information visit Whippet’s GitHub page.

Warning

The order of the columns is not relevant, however the column names must correspond to the ones indicated above. Additional columns with other names will be ignored.

Note

We recomend to consider A.number_of_pools and B.number_of_pools values that ensure that at least five cells are merged within each pseudo-bulk pool. We also suggest 10 as a minimun value of Repeat, higher values will enable better estimation of parameters to fit the resultant probabilities into beta distribution models.

Optional Configuration¶

The following parameter are optionals to be defined inside config.yaml file:

seed : 123
Only_snakepool : T
Get_Bamfiles : T

seed define a specific seed for pseudo number geration. This number influence the arrangement cells into the corresponding pseudo-bulks. Mataining the same seed ensures reproducibility of the results and prevent snakemake of overwrite completed results.
Only_snakepool is a bolean variable that if its defined as T it will force MicroExonator to skip Disovery and Quantification modules. This mode is useful for users who are only interested to find alterantive splicing events from splicing nodes that can be extracted from the annotation.
Get_Bamfiles correspond to a bolean variable that if its defined as T enable the generation of BAM files that can be used for visualization purposes.

Run¶

After setting up all the files described above, this single cell analysis module can be run by adding snakepool as target for snakemake:

snakemake -s MicroExonator.smk  --cluster-config cluster.json --cluster {cluster system params} --use-conda -k  -j {number of parallel jobs} snakepool

Note

It is allways a good idea to use -np to execute an snakemake dry-run before submiting a large set of jobs.

Unpooled quantification (optional)¶

In order to generate PSI quantification files at the single cell level (as opposed to pseudo-bulks), you can run MicroExonator with quant_unpool_single_cell as a target for snakemake. By doing this .psi.gz files will be generated at Whippet/Quant/Single_Cell/Unpooled/ folder:

snakemake -s MicroExonator.smk  --cluster-config cluster.json --cluster {cluster system params} --use-conda -k  -j {number of parallel jobs} quant_unpool_single_cell

This can enable users do run custom downstream analysis over alternative splicing quantification files generated for every cell by separated.

Warning

Only FASTQ files from cells annotated on cluster_metadata file will be processed.

Output¶

Direct results from whippet-delta for every comparison across each computational replicate can be found at Whippet/Delta/Single_Cell/. Integrated results for each comparion can be found at Whippet/Delta/Single_Cell/Sig_nodes, these resuls are structured as follow:

**all_nodes.microexons.txt**¶
Column	Description
Gene	Gene ID
Node	Node number inside the gene
Coord	Node coordinate
Strand	Plus or minus strand
Type	Node type. For more information visit Whippet’s GitHub page.
Psi_A.mean	Mean PSI values for group `cluster A` across computational replicates.
Psi_B.mean	Mean PSI values for group `cluster B` across computational replicates.
DeltaPsi.mean	Mean DeltaPsi values obtained across computationa replicates.
DeltaPsi.sd	Standar deviation of DeltaPsi values obtained across computationa replicates.
Probability.mean	Mean probability of differential inclusion obtained across computationa replicates.
Probability.var	Variance of probability across computationa replicates.
N.detected.reps	Number of replicates in which the differential inclusion could be assessed.
cdf.beta	p-value of being above the used defined probability threshold `cdf_t`
is.diff	Bolean variable defining wheather the node was differentially included accoding to the user-defined criteria (`min_rep`, `min_p_mean` and `min_delta`)
microexon_ID	Microexon ID based on its genomic coodinates.

Visualization¶

In order to visualize the results, you need to instruct whippet-quant to generate SAM files by incorporating the following parameter with a True value inside config.yaml:

Get_Bamfiles : T

Samfiles are further converted to BAM files and corresponding index files are genrated to enable their visualization. To generate these BAMs cluster_bams needs to be defined as an snakemake target:

snakemake -s MicroExonator.smk  --cluster-config cluster.json --cluster {cluster system params} --use-conda -k  -j {number of parallel jobs} cluster_bams

A BAM file will be generated for every cell-type defined at cluster_metadata file. Given the coodinates of differentially included splicing nodes and the correspondig BAM files, sashimi plots can be generated by using tools such as ggsashimi or IGV.