Synopsis

R-loops are transient three-stranded nucleic acids that form during transcription when the nascent RNA hybridizes to the template DNA, freeing the DNA non-template strand. There is growing evidence that R-loops play important roles in physiological processes such as control of gene expression, while also contributing to chromosomal instability and disease. It is known that R-loop formation is influenced by both the sequence and the topology of the DNA substrate, but many questions remain about how R-loops form and the 3-dimensional structures that they adopt.

Here we represent an R-loop as a word in a formal grammar called the R-loop grammar and predict R-loop formation. We train the R-loop grammar on experimental data obtained by single-molecule R-loop footprinting and sequencing (SMRF-seq). Despite not containing explicit topological information, the R-loop grammar accurately predicts R-loop formation on plasmids with varying starting topologies and outperforms previous methods in R-loop prediction.

Experimental data set:

Fasta & BED files

Dependencies:

• Python3
• matplotlib
• openpyxl

---

Installation

You can install directly from GitHub using the command below,

pip install git+https://github.com/Arsuaga-Vazquez-Lab/R-loopGrammar.git

Or you can clone the repo and install from the repository itself.

git clone https://github.com/Arsuaga-Vazquez-Lab/R-loopGrammar.git
cd R-loopGrammar
pip install .

Both of these methods will install the depedencies themselves.

Usage

After installation, you'll have access to the following programs,

• rloop-grammar-build-model
• rloop-grammar-union-models
• rloop-grammar-predict

These programs depend on a configuration file of plasmids/genes named plasmids.ini.

This file will contain the information regarding the plasmids that each program operates on, e.g.

[Plasmid1]
GeneStart = 80
GeneEnd = 1829
FastaFile = Plasmid1.fa
BEDFile = Plasmid1.bed

[Plasmid2]
GeneStart = 80
GeneEnd = 1512
FastaFile = Plasmid2.fa
BEDFile = Plasmid2.bed

This file defines the start and end of the gene inside the plasmid (0-based indexing, start inclusive and end exclusive), the fasta file location, and the BED file containing the experimental R-loop data.

The BED files are of the form,

plasmid1_rloop1 85 125
plasmid1_rloop2 67 342
...

These will contain the experimental results where the first index and last index are where an R-loop has been experimentally found inside the given plasmid.

1. To build a model collection we run the following,

rloop-grammar-build-model Collection_Plasmid1_runs_30 -c 30 -p 13 -w 4 --plasmids Plasmid1 -tp 10

• --plasmids The plasmids which to build the model from, as defined in plasmids.ini.
• -tp The training set precentage to use for the accomponing BED file for each plasmid.
• -c The number of runs to generate for each model collection.
• -w The k-mer size.
• -p The padding used for the sliding windows in the critical regions.
• -d (Optional) To duplicate a run utilizing the same seed or training set, use this option to select a model to copy from; this will override -c.

2. To union two model collections together, we then run,

rloop-grammar-union-models UnionCollection_Plasmid1_Plasmid2 -m stochastic -i Collection_Plasmid1_runs_30 Collection_Plasmid2_runs_30

• -i The two input plasmids which are used to build the union model.
• -m The method used to take the union, the current supported options are stochastic and deterministic.

3. Then we can make a prediction.

rloop-grammar-predict UnionCollection_Plasmid1_Plasmid2_predict_on_Plasmid3 -i UnionCollection_Plasmid1_Plasmid2 --plasmids Plasmid3

• -i The input model used to build the prediction.
• -p The plasmid used to predict upon.

4. And finally we can graph the prediction.

rloop-grammar-graph-prediction UnionCollection_Plasmid1_Plasmid2_predict_on_Plasmid3 -n Prediction_Plasmid3

• -n The name displayed on the graph.

Reproduce model data

If you would like to reproduce the model data found here, download the zip file. We will be using the random seed for each run to recreate the data utilizing the -d duplicate option when building the model.

First we need to patition the bed-files utilizing the training data seed. The make_training.py script will partition the BED files into three sepearate sets utilizing the seed file. We will be using the first partition to build our model.

The plasmids.ini will expect the BED files to be located in a folder called bed-files, also with the FASTA files in fasta-files. You can find the plasmids.ini file also in the training-data.zip.

The following Makefile can be used to then recreate the model.

all: models union predict

models:
	rloop-grammar-build-model new_Collection_pFC53_SUPERCOILEDCR_runs_30 -c 30 -p 13 -w 4 --plasmids pFC53_SUPERCOILEDCR_training -tp 10 -d original_Collection_pFC53_SUPERCOILEDCR_runs_30
	rloop-grammar-build-model new_Collection_pFC8_SUPERCOILEDCR_runs_30 -c 30 -p 13 -w 4 --plasmids pFC8_SUPERCOILEDCR_training -tp 10 -d original_Collection_pFC8_SUPERCOILEDCR_runs_30
	rloop-grammar-build-model new_Collection_pFC53_GYRASECR_runs_30 -c 30 -p 13 -w 4 --plasmids pFC53_GYRASECR_training -tp 10 -d original_Collection_pFC53_GYRASECR_runs_30
	rloop-grammar-build-model new_Collection_pFC8_GYRASECR_runs_30 -c 30 -p 13 -w 4 --plasmids pFC8_GYRASECR_training -tp 10 -d original_Collection_pFC8_GYRASECR_runs_30
	rloop-grammar-build-model new_Collection_pFC53_LINEARIZED_runs_30 -c 30 -p 13 -w 4 --plasmids pFC53_LINEARIZED_training -tp 10 -d original_Collection_pFC53_LINEARIZED_runs_30
	rloop-grammar-build-model new_Collection_pFC8_LINEARIZED_runs_30 -c 30 -p 13 -w 4 --plasmids pFC8_LINEARIZED_training -tp 10 -d original_Collection_pFC8_LINEARIZED_runs_30

union:
	rloop-grammar-union-models new_UnionCollection_SUPERCOILEDCR_runs_30 -m stochastic -i new_Collection_pFC53_SUPERCOILEDCR_runs_30 new_Collection_pFC8_SUPERCOILEDCR_runs_30
	rloop-grammar-union-models new_UnionCollection_GYRASECR_runs_30 -m stochastic -i new_Collection_pFC53_GYRASECR_runs_30 new_Collection_pFC8_GYRASECR_runs_30
	rloop-grammar-union-models new_UnionCollection_LINEARIZED_runs_30 -m stochastic -i new_Collection_pFC53_LINEARIZED_runs_30 new_Collection_pFC8_LINEARIZED_runs_30

predict:
	rloop-grammar-predict new_UnionCollection_SUPERCOILEDCR_predict_on_pFC53 -i new_UnionCollection_SUPERCOILEDCR_runs_30 --plasmids pFC53_SUPERCOILEDCR
	rloop-grammar-predict new_UnionCollection_SUPERCOILEDCR_predict_on_pFC8 -i new_UnionCollection_SUPERCOILEDCR_runs_30 --plasmids pFC8_SUPERCOILEDCR
	rloop-grammar-predict new_UnionCollection_GYRASECR_predict_on_pFC53 -i new_UnionCollection_GYRASECR_runs_30 --plasmids pFC53_GYRASECR
	rloop-grammar-predict new_UnionCollection_GYRASECR_predict_on_pFC8 -i new_UnionCollection_GYRASECR_runs_30 --plasmids pFC8_GYRASECR
	rloop-grammar-predict new_UnionCollection_LINEARIZED_predict_on_pFC53 -i new_UnionCollection_LINEARIZED_runs_30 --plasmids pFC53_LINEARIZED
	rloop-grammar-predict new_UnionCollection_LINEARIZED_predict_on_pFC8 -i new_UnionCollection_LINEARIZED_runs_30 --plasmids pFC8_LINEARIZED

.PHONY: models union predict

make all

References

1. Stolz R, Sulthana S, Hartono SR, Malig M, Benham CJ, Chedin F. Interplay between DNA sequence and negative superhelicity drives R-loop structures. Proc Natl Acad Sci U S A. 2019 Mar 26;116(13):6260-6269. doi: 10.1073/pnas.1819476116. Epub 2019 Mar 8. PMID: 30850542; PMCID: PMC6442632.
2. Jonoska N, Obatake N, Poznanovik S, Price C, Riehl M, Vazquez M. (2021). Modeling RNA:DNA Hybrids with Formal Grammars. 10.1007/978-3-030-57129-0_3.
3. Ferrari M, Poznanovik S, Riehl M, Lusk J, Hartono S, Gonzalez G, Chedin F, Vazquez M, Jonoska N. (Submitted). The R-loop Grammar predicts R-loop formation under different topological constraints.