LIQUORICE logo

Introduction

LIQUORICE is a command-line-tool and python package for bias correction and quantification of changes in sequencing coverage around regions of interest whole-genome sequencing datasets of cell-free DNA. LIQUORICE can be used to detect and quantify tissue or cancer-specific epigenetic signatures in liquid biopsies. This allows accurate quantification of the fraction of tumor-derived cell-free DNA, as demonstrated in our publication in Nature Communications. As a tool, LIQUORICE has recently been published in Bioinformatics Advances.

Motivation

The fragmentation of cell-free DNA (cfDNA) obtained from liquid biopsies is non-random, and contains information about the epigenetic state of its cell-of-origin. DNA in nucleosomes is protected from cleavage, resulting in an increase in sequencing coverage in genomic regions that are occupied by nucleosomes. If a set of genomic regions is specifically accessible in a cell-, tissue- or cancer type of interest, an observed decrease in sequencing coverage around these regions can indicate that cells of this type are present in the organism, and are releasing cfDNA into the bloodstream.

Potential application areas

LIQUORICE could be useful for a variety of cfDNA-related analysis tasks:

  • Cancer detection, classification, monitoring and survival prognosis (as demonstrated in our recent publication)

  • Classification of cancers of unknown primary

  • Location of metastasis

  • Tissue damage monitoring for transplantations, heart attack, and other diseases

What LIQUORICE does

LIQUORICE addresses two problems regarding the quantification of epigenetic signatures from a sequencing coverage signal:

  • Coverage is influenced by bias-factors, such as GC-content, mappability and other factors related to sequence composition

  • The coverage signal needs to be summarized over all regions in a region-set, and needs to be quantified as a number that is robust and comparable between samples

In a first step, LIQUORICE trains a bias-model, which learns the association between bias-factors and coverage. Then, the trained model is used to correct the coverage signal in the regions of interest. Finally, the signal is aggregated between regions and quantified by fitting a model to the data. LIQUORICE produces tables as well as plots that allow for the visual inspection of biases and of the coverage signal. Furthermore, a summary tool is included, which allows for the convenient comparison of signals between samples and region-sets.

Here are a few examples of LIQUORICE’s output:

Bias correction example Model fitting example Overlay plot / Multi-sample summary example

Region-sets

In the context of LIQUORICE, we refer to a region-set as a set of genomic regions with similar properties. This could be, e.g., enhancers or DNase I hypersensitivity sites that are specific for a cell-type, cancer-type, or tissue. Useful ressources to obtain such region-sets include http://dnase.genome.duke.edu/ and ENCODE.

Here is an example on how we obtained liver-specific DNase I hypersensitivity sites from the http://dnase.genome.duke.edu/celltype.php interface. First, we required all non-liver cells to be “closed”, and we required hepatocytes to be “open” at a given DHS:

DHS selection example - 1

Then, we selected two clusters that have an adequate tissue-specific accessibility pattern:

http://dnase.genome.duke.edu/clusterDetail.php?clusterID=1066

DHS selection example - 2

and http://dnase.genome.duke.edu/clusterDetail.php?clusterID=1115

DHS selection example - 3

We downloaded the .bed files (with all regions) from these clusters, and finally, we concatenated the downloaded files, merging the DHSs of these two clusters in order to get our final liver-specific DHS set. Note that the regions on the database are stored in hg19 coordinates - you can use LiftOver to convert the data to hg38.

How to use LIQUORICE

The easiest way to use LIQUORICE is via its command-line interface. Also check out the convenient summary tool if you are analyzing multiple samples or region-sets.

If you require more control about the steps performed by LIQUORICE, check out the LIQUORICE python package - you can start by looking at the function-level documentation and source code of the liquorice.LIQUORICE module as well as the the liquorice.utils.Workflows module.

More details on parameters, test examples and usage examples can be found here

Contact

If you have any questions about LIQUORICE and how to apply it to your data, create an issue on github or contact peter.peneder@ccri.at - we are happy to hear from you.

Source code on Github

The github repository of LIQUORICE can be found at https://github.com/epigen/liquorice.

Installation

conda

LIQUORICE can easily be installed via conda (or mamba , a faster alternative to conda). We recommend installing LIQUORICE in its own conda environment to keep dependencies clean:

# to install on Linux
conda create -n LIQUORICE -c bioconda -c conda-forge liquorice ray-core

# to install on macOS
# conda create -n LIQUORICE -c bioconda -c conda-forge liquorice

# to activate the environment
conda activate LIQUORICE # or: 'source activate LIQUORICE' for older conda versions

# to run
LIQUORICE <flags and arguments>

‘ray-core’ is an optional dependency of LIQUORICE that helps it run faster on multiple cores. It is not possible to install it via conda on macOS, so installing it can either be skipped (find details about an alternative parallelization approach here) or it can be installed as described here (e.g. pip install -U ray==1.1.0). LIQUORICE has been tested for ray version 1.1.0.

docker

You can also use the LIQUORICE docker image which is available here. To pull, use

docker pull peneder/liquorice

Citation

If you use LIQUORICE in any published work, please cite:

Peneder, P., Bock, C. & Tomazou, E. M. (2022). LIQUORICE: detection of epigenetic signatures in liquid biopsies based on whole-genome sequencing data. Bioinformatics Advances, 2(1), vbac017.

Peneder, P., Stütz, A. M., Surdez, D., Krumbholz, M., Semper, S., Chicard, M., … & Tomazou, E. M. (2021). Multimodal analysis of cell-free DNA whole-genome sequencing for pediatric cancers with low mutational burden. Nature communications, 12(1), 1-16.