BGlab Smart-Seq2 preprocessing toolkit

Introduction

smqpp is the preprocessing pipeline for Smart-Seq2 data, specifically for datasets generated from Gottgens lab. The QC part of code was adpoted from bglab package developed by Wajid Jawaid.

The package contains the following steps:

1. Preanalysis

   • generate_feature_table: If gene feature table not available then this can be generate using this function
   • read_in_files: Read in count and QC inputs and format them into anndata object
   • reformat_meta: Reformat metatable to keep all versions consistent (Due to different versions of metadata spread sheet from google drive)
   • smartseq_qc: bglab equivalent quality control
   • normalise_data: Data normalisation using DESeq2 method
   • quantile_norm: Quantile normalisation
   • quantile_norm_log: Log quantile normalisation
   • downsampling_norm: Downsampling normalisation (Not recommanded for TenX as it will shrink more counts to 0)
   • tech_var: Highly variable gene (HVG) calculation using Brennecke et. al method
   • plot_tech_var: Plot the HVG prediction
   • detect_outlier_cells: filter out cells that effect the selection of HVGs

2. Differential expression analysis

   • plot_ma: MAplot for rank_genes_group from Scanpy and select significant genes with high confidence

3. Pseudotime time analysis

   • GeneExp_LLR_test: Likelihood ratio test to select genes that differentially expressed along pseudotime. Linear models were fitted between log norm exp and smoothed PT by applying natural spline.
   • plot_genes_along_pt: Plotting out gene expression pattern along PT. Gene exp was smoothed using Guassian filter.

4. Projection

   • quick_neighbors: Neighbors calculation adpoted from scanpy. Two constraints applied: 1) reference cells only allow neighbors between themselves; 2) new cells only allow neighbors with reference cells
   • quick_umap: Similar to the ingest function in scanpy. Umap was calculated using umap python package. Parameters used as scanpy defaults.
   • quick_umap_proj: Projection of new data onto reference data

5. 3d plots

   • plot_3d: Generate 3d plots from anndata object as the projection='3d' function does not work properly in the latest scanpy due to matplotlib issues

6. Pathway analysis:

   • pathway_score_cal: Calculate geometric mean for each terms in the databse for each cell, which can be used to color the defined layout
   • pathway_analysis: Calculate the enriched database terms for a given gene set using hypergeometric test.

Installation

smqpp depends on numpy, matplotlib, pandas, anndata, scipy and statsmodels. The package is available on pip and can be easily installed as follows:

pip install smqpp

or

download the file from github using git clone
tar zxvf smqpp
cd smqpp
pip install .

Usage and Documentation

The smqpp should be fairly simple to use and it is based on Scanpy's AnnData object:

import smqpp
smqpp.read_in_files(...)

Example Notebooks

Examples can be found in the following folders:

1. GSK analysis
2. Patel study

Contact

If there are any issues, please contact xw251@cam.ac.uk.