What is the SpliceDecoder?

• Splice decoder provides functional annotation for your differential splicing events (DSEs)
• The functional annotation contains NMD probability, alterations in functional domains (such as DNA binding, motif, regions, protein domain, and so on), CDS/UTR alterations, and effect score
• The effect score can be used to prioritize and choose the most representative functional consequences of your DSEs
• Currently, SpliceDecoder supports hg38 and mm10 genome

Workflow overview

1. Generate All Possible Splicing Cases (Processing input) : This step makes proper format of input data from the output of event-based splicing tools
2. Map Splicing Cases (Mapping DSEs and ORFs): This step explores the given transcriptome (.GTF) to find Ref-TX (Reference transcript, it contains perfectly matched exon structure for the given DSE) and assign the best three open reading frames (ORFs)
3. Simulate Splicing Events (Simulation): Based on the Ref-TXs and their ORFs, this step perform simulation of alternative splicing (e.g., if the Ref-TX has exon inclusion (EI) form, this step makes a simulated transcript (Sim-TX) with exon skipped (ES) form)
4. Functional Annotation (Annotation): Based on the Uniprot DB, SpliceDecoder assigns known functional domains and estimates functional changes between Ref-TX and Sim-TX
5. DSEs with Effect Score (Scoring): SpliceDecoder assigns an effect score to each DSE based on multiple biological factors, enabling prioritization of your DSEs

Quick start (conda is required)

• SpliceDecoder can be downloaded from https://github.com/hyeon9/SpliceDecoder/

• Install SpliceDecoder by using the install script

  cd ./SpliceDecoder && bash install.sh
  

• To perform a test run, you can use the provided toy_data

• You can make a toy configuration file through an interactive way You can find more details here

  cd code/
  bash Make_config.sh toy
  

• If you successfully created Your_toy.config, you can run SpliceDecoder

• The steps are intended to be executed in order, so it is recommended to use all

  bash Main.sh all ${Your_toy.config}
  

• If your test run with the toy data finishes successfully, you will see the following output files (except for the Effect_score.tsv)

• To run with your own data, create a configuration file using the Make_config.sh or modifying example.config

  cd code/
  bash Make_config.sh
  bash Main.sh all ${Your.config}
  
  OR
  
  vi example.config
  mv example.config ${Your.config}
  

• Then, use this command to submit your job if you are using SLURM

• You can specify #SBATCH options such as -c 10 AND --mem=40G

  sbatch Main.sh {Make_input | DS_mapping | ORF_mapping | Simulation | Scoring | all} ${Your.config}
  

• If needed, you can run a specific step by selecting one of the following: Make_input, DS_mapping, ORF_mapping, Simulation and Scoring

  bash Main.sh {Make_input | DS_mapping | ORF_mapping | Simulation | Scoring} ${Your.config}
  

• If you want to annotate a transcript-centric data you can find more details here

Guide for making config file

• Make_config.sh will ask..

  ? Specify your config file name (e.g. HGjob)
  > You just need to specify your config file
  
  ? Enter the path of SpliceDecoder (e.g. /User/usr/Tool/SpliceDecoder-main/)
  > You just need to specify the install path of SpliceDecoder
  
  ? Enter your working directory (e.g. /User/usr/Tool/SpliceDecoder-main/project1)
  > You just need to specify your new working directory
  
  ? Enter your rMATS output path (e.g. /User/usr/Tool/SpliceDecoder-main/toy_data)
  > You just need to specify the rMATS output path
  
  ? Enter your target gene list (e.g. /User/usr/Tool/SpliceDecoder-main/target_genes.tsv)
  > You just need to provide interesting gene list, or enter 'all' if you don’t have one
  > SpliceDeocder will only consider there genes
  
  ? Enter your GTF file that you used in rMATS with its full path (e.g. /User/usr/Tool/SpliceDecoder-main/toy_data/toy.gtf or /User/usr/Tool/SpliceDecoder-main/toy_data/*.gtf)
  > You just need to specify the full path + GTFfile
  
  ? Do you want to calculate the effect score? [yes/no]
  > Simply type yes or no. If you type "yes", SpliceDecoder will ask TPM matrix or bamfile path to calculate the effect score
  
  ? Enter your TPM matrix with full path (e.g. /User/usr/Tool/SpliceDecoder-main/toy_data/tpm.tsv or N)
  > Specify the full path to your TPM matrix, or enter 'N' if you don’t have one
  
  ? Enter your bamlist which should contains bamfile with their full path in each line (e.g. /User/usr/Tool/SpliceDecoder-main/toy_data/bam_list.txt or N)
  > If you don’t have a TPM matrix, specify the full path to your BAM list file, or enter 'N'
  
  ? Enter a species of your data (e.g. human or mouse)
  > You just need to specify the species of your data
  
  ? Enter a type of GTF (e.g., SR (GENCODE GTF) or LR (Custom GTF) )
  > You just need to specify the type of your GTF
  
  ? Specify a NMD definition method (e.g., default (55rule) or advanced) )
  > You just need to select one either 'default' or 'advanced'
  
  ? Enter a FDR cut off for your rMATS (float [0-1], default 0.05)
  > Specify rMATS FDR cut off
  
  ? Enter a |dPSI| cut off for your rMATS (float [0-1], default 0.1)
  > Specify rMATS FDR cut off
  
  ? Enter a number of cpu in spliceDecoder job (int [0-?])
  > Specify a number of cpu will be used in your job
  

• You can reuse a pre-existing config file by copying it:

  cp ${existing_config} project2.config
  

• Then, update the following fields in the new config: input, Your_GTF, and Your_rMATS

Outputs

├── table/
│   ├── *_w_Pfam.txt: Assigned domain information of simulated transcripts (Sim-TXs)
│   └── *_wo_Pfam.txt: Assigned domain information of reference transcripts (Ref-TXs)
├── result/
│   ├── *Main_table.tsv: description
│   ├── *Domain_alt.tsv: description
│   ├── *NMD.tsv: description
│   └── Effect_score.tsv: description
├── figure/
│   ├── mapping_rate.pdf: Mapping rates for each splicing type
│   ├── mat_tx_numbers.pdf: Distribution of Ref-TX for each splicing type
│   ├── splicing_categories_stacked_plot.pdf: description
│   ├── merged_stacked_plot.pdf: description
│   ├── Summary.html: HTML file to make summary pages (pdf_1_page_1.png, pdf_2_page_2.png, pdf_3_page_3.png, and pdf_4_page_4.png)
│   └── consequence: Output directory of visualization script
├── AF2/: Contains AlphaFold2 input (amino acid FASTA)
├── temp/: Contains all intermediate files
├── post_input/: Contains files used in downstream analyiss e.g., visualization and 3D structure generation
├── mapping.stats: Mappeing rates for each splicing type
└── SD.log: The log file

• Example of summary HTML

Details of Outputs

Example of the Effect_score.tsv

Key Metrics

• LongID: DS event ID
• Gene symbol: Gene symbol
• Reference_transcript: Matched Transcript (==Ref_TX)
• Simulated_event: Simulated event (ES = Exon skipping, EI = Exon inclusion, SI = Skipped intron, RI = Retained intron, Can A3/5SS = canonical 3/5' splice site, Alt A3/5SS = alternative 3/5' splice site)
• Effect_Score: A score to prioritize your DS events [0,2]
• Domain_change_rate: Average rate of domain changes in Sim-TX compared to Ref-TX [0,1]
• Probability_of_NMD: NMD (-1), PTC removal (1), No NMD related event (0)
• Functional_class: It contains the following functional classes: GoD (Gain of Domain), LoD (Loss of Domain), NMD, CDS_alts, and UTR_alts

Supplementary Metrics

• Delta_PSI: PSI difference (group2 - group1) [-1,1]
• Transcript_usage: Proportion of expression of reference transcript for each gene [0,1]
• ORF: Used ORF (This file only contains pORF1 which has the highest coding potential)
• AUG (Ref-Sim): Start codon position on the Ref TX and Sim TX (Ref-Sim)
• Stop_codon (Ref-Sim): Stop codon position on the Ref TX and Sim TX (Ref-Sim)
• Nucleotide_difference: Coding sequence length difference (Ref TX - Sim TX)
• 5'UTR_difference: 5' UTR length difference (Ref TX - Sim TX)
• 3'UTR_difference: 3' UTR length difference (Ref TX - Sim TX)
• Domain_integrity: (Sim_domain_length / Ref_domain_length) * 100 [0,inf]
• Length_of_simulated_tx_domain: Total domain length of Sim TX
• Length_of_referece_tx_domain: Total domain length of Ref TX
• rMATS_FDR(-log10): -Log10 scale FDR, it came from rMATS

---

Example of the Domain_alt.tsv

Key Metrics

• DS-TX pair ID: It contains, in order Long_ID, Ref-TX ID, and simulated event type
• ORF priority: A priority of the used reading frame in simulation
• Domain information: A name of altered domain by the simulated alternative splicing event
• Functional_change_ratio (∆L): A difference of functional change ratio for simulated alternative splicing
• Change direction: It indicates whether the altered domain is a gain (1) or a loss of domain (-1)

---

Example of the NMD.tsv

Key Metrics

• LongID: Contains, in order Long_ID, Ref-TX ID, and simulated event type
• AUG: A relative position of AUG on the given transcript (Ref or Sim)
• pORF: A priority of the used reading frame in simulation
• distance(last_exon_junction-stop): Distance between last exon-exon junction and stop codon (calculated by last_exon_junction - stop)
• total_domain_length: Total domain length of the given transcript (Ref of Sim)
• key(Ref/Sim): A type of transcript (Ref of Sim)
• NMD_possibility: Indicates the possibility of NMD. In default mode, values are HIGH (55nt) or No. In advanced mode, values can be HIGH (55nt), INTERMEDIATE (Long-exon), INTERMEDIATE (Start-proximal), LOW (less 55nt), or No. Only events tagged as HIGH are considered NMD-associated events
• contain_PTC: Indicates whether the given transcript contains PTC (Y) or not (N)

Visualize your alternative splicing simulation

• Based on your Main_table file, you can pcik ceratin DS event to visualize it using this code

  conda activate spliceDecoder
  python code/02-3_v3_Draw_consequence.py \
         --input ${working directory} \
         --splicing_event RI \
         --gene MYLK2 \
         --sim_splicing_event RI \
         --transcript ENSMUST00000028970.7
  python code/02-3_v3_Draw_consequence.py -h  # You can get more details
  

• If you want to remove some information in figure space, using ri option (all categories should be separated by space)

  python code/02-3_v3_Draw_consequence.py \
         --input ${working directory} \
         --splicing_event A3SS \
         --gene MYLK2 \
         --sim_splicing_event Can_A3SS \
         --transcript ENSMUST00000195957.4 \
         -ri proteome chain
  

• All figures will be saved at ${input}/figure/consequence/

Create a 3D Protein structure based on simulated

• You can use Make_aa_fa.py to extract amino acid sequences from your interesting targets

• This function requires the Effect_score.tsv, Toy data is not eligible for this function

• You can find the ${input} and ${Main} in your .config file

  conda activate spliceDecoder
  python code/Make_aa_fa.py \
         -i ${input} \
         -r human \
         -t ENST00000438015.6 \
         -e ES \
         -d ${Main}
  

• You can copy and paste the amino acid sequences to the Alphafold server (https://alphafoldserver.com) as input

Please cite this article if you use SpliceDecoder in your research

• https://doi.org/10.1101/2025.10.01.679902