sinomem / qctreecnv

Powerful CNV calls filtering pipeline. Focused on fixed loci calling, before visual inspection.

QCtreeCNV

Introduction

This R package implement a powerful CNV filtering pipeline as exemplified in the following figure. It is part of a protocol with a companion repository, https://github.com/SinomeM/IBPcnv.

If you use thih software please cite https://doi.org/10.1002/cpz1.621

The filtering pipeline

The filtering pipeline is implemented as a decision making tree, a series of consecutive steps that separate good putative candidates form bad ones. This scheme describe the process.

Install

You can install the package via:

devtools::install_github("sinomem/QCtreeCNV")

Details

See the package vignettes for more details. Assuming the package is installed in your system you can do:

browseVignettes("QCtreeCNV")

qctreecnv's Issues

add the possibility to parallelize the slower steps

add merge CNVRs to the fast implementation?

check when `dt` copies are made and if it is really necessary

`check_inputs()` is defined but not integrated in `qc_pre()`

ADD a function that run the entire pipeline w/ all required checks

can take inspiration from the latest R script in the 30 loci analysis

select_filter() should create the length column if not already present

try vectorize some for loops

e.g. the one sample wise in extractMetrics ?

Write some minimal formal documentation for each exported function + the package itself

`cnvr_fast()` need to make it optional to check for the possibility of just one cluster

add the possibility to skip individual steps in `qctree()`

e.g. one may be interested in searching small CNVRs only, steps 2 & 4

split qc_tree() into two funtions, one to just check the inputs (+ preprocessing) the other to run the pipelne ?

I think the first would be much more useful.
The pipeline is designed to be run one step at the time at its "core"

compute CNVRs per locus separately?

It would make sense in this context. Maybe it is something that can be integrated in the next version of computeCNVRs()

Fix qctree_pre()

remove duplicates + the extractMetrics() call

Check better the behaviour of qctree()

All functions needs some initial check to be sure the inputs are correct

add support for the use of snppos file in `extractRDS()`

as "promised" i the vignettes

numsnp column is necessary in select_stitch() funciton but it is not listed in the documentation

Integrate initail QC and stitching as a function in the package

Basically adapt Andres code to the data format used here

Integrate CNVgears functions

There are several (mostly helper/internal) functions copied from CNVgears, at the moment is the simplest thing. In the next future it would be much better to have them in a single package so development and maintenance is easier

integrate TABIX