Inquiry About Integrating RBP Binding and Knockdown RNA-seq in Splicekit

[santataRU]

Dear Splicekit Developers,

I was wondering whether Splicekit supports the integration of RBP binding data (e.g., CLIP-seq peaks) with RNA-seq data from RBP knockdown experiments to investigate differential alternative splicing. Specifically, can Splicekit be used to assess whether RBP binding near splice sites directly influences their usage upon knockdown?

I’m aware that rMAPS2 offers similar functionality, but its source code is not publicly available and appears not to have been updated since 2020.

Thank you in advance for your time and help.

Best regards,
Xiao

[grexor]

Dear Xiao,

Thanks for using splicekit and for reaching out!

Indeed, splicekit calls scanRBP to estimate protein binding and then uses a bootstrap approach to estimate significance of binding in splice site regulation.

You can either use PWD for various proteins to estimate binding (check scanRBP docs please) or use *CLIP peaks (BED files) directly to provide the signal.

This part of code in core/motifs.py does the scanRBP calling:

    if config.clip not in [None, ""]:
        # user provided bedGraph with CLIP peaks?
        print(f"scanRBP {fasta_fname} --output_folder results/motifs/scanRBP/data -nonzero -protein {config.protein} -clip {config.clip}")
        os.system(f"scanRBP {fasta_fname} --output_folder results/motifs/scanRBP/data -nonzero -protein {config.protein} -clip {config.clip}")

Notice the -clip flag for scanRBP. You can provide a BED file (.gz or without) in the splicekit.config file clip parameter. Please try it out and report here if you encounter any problems.

Thanks,
Gregor

[santataRU]

Dear Gregor,

Thank you very much for your suggestions! I have PAR-CLIP data and used PARalyzer to call clusters (i.e., RBP binding sites). My cluster file looks like this:

Chromosome	Strand	ClusterStart	ClusterEnd	ClusterID	ClusterSequence	ReadCount	ModeLocation	ModeScore	ConversionLocationCount	ConversionEventCount	NonConversionEventCount
chr1	+	633972	634074	G1.1	ACACCAACCACCCAACTATCTATAAACCTAGCCATGGCCATCCCCTTATGAGCGGGCGCAGTGATTATAGGCTTTCGCTCTAAGATTAAAAATGCCCTAGCCC	25	633972	0.994064805	13	23	299
chr1	+	1043889	1043935	G2.1	GGTGTGAGCACCCCCCGCCCGGCCCCGTGTGTGGCAGCGACGGTGTC	7	1043904	0.961357292	2	2	41
chr1	+	1047087	1047109	G3.1	GTGCCGGGGACCCCACGCCTAAC	7	1047100	0.639514255	2	6	8
chr1	+	1047227	1047264	G4.1	CTTCCCTTGCACCCTTGTGGCTTGCTGCTGGGGCCTGT	11	1047253	0.625050639	5	13	98
chr1	+	1047274	1047320	G4.2	AGCCTGGATGCCAGGCAGATGCCAGGCAGGGCCTCACTGTACCTCCC	14	1047285	0.940508059	4	13	50
chr1	+	1050380	1050416	G5.1	CTGTGGGGAGGGGACAGCAAAGACACCCCGACTCCCC	8	1050397	0.905416546	2	5	15
chr1	+	1055056	1055110	G6.1	CTTGCCTGAGTGTTGGCCGGAGGGACTGCTGGCCCGGCCTCCCTTCCGTCCAGGC	17	1055056	0.773337949	8	14	123
chr1	+	1351780	1351804	G7.1	AGGGAGGGAGGAAGCCTGTCCCTGC	5	1351788	0.90139684	1	5	10
chr1	+	1353932	1353960	G8.1	TGGGAGGAGAGGAGGCTGAGGTCCCCGCT	5	1353960	0.799630512	3	5	15
chr1	+	1477229	1477253	G9.1	ATCCGTGTATCCTACACCTGCTCTC	5	1477253	0.642258187	6	8	32
chr1	+	1477604	1477640	G10.1	GAGGGCATCTGGTCGTCCTGAGGGAGGGGGTCTTCTT	7	1477626	0.725383493	6	7	47

I know that a BED file requires at minimum three columns: chrom, chromStart, and chromEnd. For scanRBP, do you need any additional columns—such as ReadCount or ModeScore—to perform the analysis?

ReadCount: number of reads overlapping the cluster by at least one nucleotide

ModeScore: ratio of highest signal to (signal + background).

Thank you very much for developing these valuable tools!

Best regards,
Xiao

[grexor]

Cool, please try putting ReadCount for the BED value, should work. Keep in mind the CLIP integration was never thoroughly tested so I envision some problems, that's why its great you are testing and running it, we can fix potential problems together.

Best, Gregor

[santataRU]

Thanks. Which column in the BED file should I use to include the “ReadCount”? Are you suggesting I create a four-column BED file including this value?

Best, Xiao

[grexor]

Try to use a simple bedGraph format instead (chr, start, stop, value), negative values for - strand: https://genome.ucsc.edu/goldenpath/help/bedgraph.html

[santataRU]

Hi Gregor,

Thank you for uploading the YouTube video introducing the SpliceKit tool (https://www.youtube.com/watch?v=P1m73usZ3lc)! It was very helpful.

If I understood correctly, after identifying differential junction usage between two conditions (e.g., control vs. treatment) using SpliceKit, the scanRBP module only searches for CLIP-seq peaks near donor sites, but not acceptor sites. Does this mean that scanRBP would overlook RNA-binding proteins (RBPs) that primarily bind near acceptor splice sites?

Additionally, based on the video, it seems that the CLIP-seq peaks are only considered if they fall within ±80 nt of the donor site. Does this imply that peaks located further away (e.g., -500 nt or +500 nt) will be ignored by scanRBP?

Best regards,
Xiao

[santataRU]

Dear Xiao,

Thanks for using splicekit and for reaching out!

Indeed, splicekit calls scanRBP to estimate protein binding and then uses a bootstrap approach to estimate significance of binding in splice site regulation.

You can either use PWD for various proteins to estimate binding (check scanRBP docs please) or use *CLIP peaks (BED files) directly to provide the signal.

This part of code in core/motifs.py does the scanRBP calling:

    if config.clip not in [None, ""]:
        # user provided bedGraph with CLIP peaks?
        print(f"scanRBP {fasta_fname} --output_folder results/motifs/scanRBP/data -nonzero -protein {config.protein} -clip {config.clip}")
        os.system(f"scanRBP {fasta_fname} --output_folder results/motifs/scanRBP/data -nonzero -protein {config.protein} -clip {config.clip}")

Notice the -clip flag for scanRBP. You can provide a BED file (.gz or without) in the splicekit.config file clip parameter. Please try it out and report here if you encounter any problems.

Thanks, Gregor

Dear Gregor,

I noticed that the splicekit.config file provided in the example (https://github.com/bedapub/splicekit/blob/main/datasets/GSE221868/splicekit.config) does not include a clip parameter.

I would appreciate any guidance you can provide to set up the splicekit.config file to integrate splicekit analysis with CLIP-seq peaks.

Best regards,
Xiao

[grexor]

Hi Xiao,

splicekit searches for RBP binding around donor and acceptor sites. The region of search can be defined in the motifs.py script, but good point to add it as a parameter also (maybe you could do some development there?).

For the clip parameter, it should work, but you are right, we should add docs and examples for this. Also please check the motifs.py script, its all there.

Hope this helps,
Gregor

[santataRU]

Hi Gregor,

Since SpliceKit includes rMATS analysis, I was wondering — if I already have rMATS results, can I directly run ScanRBP using a CLIP-seq cluster bedGraph file along with those rMATS results for peak enrichment analysis?

Best regards,
Xiao

[santataRU]

Hi Gregor,

I’ve successfully installed splicekit (v0.7) and am eager to start testing it. Following your instructions, I noticed the following lines in motif.py:

if config.clip not in [None, ""]:
                # user provided bedGraph with CLIP peaks?
                print(f"scanRBP {fasta_fname} --output_folder results/motifs/scanRBP/data -nonzero -protein {config.protein} -clip {config.clip}")
                os.system(f"scanRBP {fasta_fname} --output_folder results/motifs/scanRBP/data -nonzero -protein {config.protein} -clip {config.clip}")

I have a couple of questions regarding the splicekit.config file setup:

1. I see the following line referencing DEXSeq scripts:

# dexseq (path to scripts)
dexseq_scripts = ""

However, this is not present in the example config files. Is installing DEXSeq required for running splicekit?

2. I would like to analyze CLIP-seq clusters (in bedGraph format), I understand that I should modify the config file to reflect the use of CLIP-seq clusters instead of RBP motifs. I assume I need to modify these lines below. Could you please provide an example configuration for enabling this analysis?

# scanRBP
scanRBP = False # run scanRBP on this dataset
protein = "TARDBP.K562.00"
protein_label = "tdp43" # used in file names and titles

Thank you very much for your help.

Best regards,

Xiao

[grexor]

Hi Xiao,

Thanks & sorry for the delayed reply. By default, splicekit will run edgeR (so you can leave out the dexseq_scripts parameter for now). For clip, please try to provide a bedGraph file (see above, -clip parameter for scanRBP).

Hope this helps,
Gregor

[santataRU]

Hi Gregor,

Thank you! Sorry, I'm a little confused — did you mean that I should use the following format?

# scanRBP
scanRBP = CLIP
protein = "Name_of_Protein"
protein_label = "bedGraph_file_name" 

If I provide the path to DEXSeq, does that mean SpliceKit will switch from differential junction usage analysis to differential exon usage analysis? If DEXSeq is used, can scanRBP still perform CLIP-seq peak enrichment analysis in the downstream analysis?

[grexor]

Sorry its a little cryptic but will document it better. For DEXseq do not worry and do not provide any parameters, the def. is edgeR. For CLIP, simply provide a clip= parameter (+ scanRBP = True) in the config file and it should pass it on to scanRBP. Hope this helps, Gregor

[santataRU]

Hi Gregor,

I installed Splicekit and set up my samples.tab and splicekit.config files. I would like to run ./run_snakemake_local.sh --configfile config.yaml, but I couldn’t find either run_snakemake_local.sh or config.yaml.

Could you please advise me on where to locate these files?

Thanks,
Xiao

(splicekit) bieniaszlab@pauls-imac-2 splicekit % pwd
/opt/anaconda3/envs/splicekit/lib/python3.12/site-packages/splicekit
(splicekit) bieniaszlab@pauls-imac-2 splicekit % ls
init.py clusterlogfc core judge report splicekit.config version
pycache config folders.setup june samples.tab splicekit.config.template
(splicekit) bieniaszlab@pauls-imac-2 splicekit % snakemake --version
9.9.0
(splicekit) bieniaszlab@pauls-imac-2 splicekit % which splicekit
/opt/anaconda3/envs/splicekit/bin/splicekit

[grexor]

Hi Xiao, good point, for now the file is not included in the pip install, but you can find it in the repo, its just a wrapper for the snakemake command: https://github.com/bedapub/splicekit/blob/main/run_snakemake_local.sh

We will/should :) include the run scripts as entry points in the next pip release.

Hope this helps, Gregor