bio-splicing-quantification

name: bio-splicing-quantification description: Quantifies alternative splicing events (PSI/percent spliced in) from RNA-seq using SUPPA2 from transcript TPM or rMATS-turbo from BAM files. Calculates inclusion levels for skipped exons, alternative splice sites, mutually exclusive exons, and retained introns. Use when measuring splice site usage or isoform ratios from RNA-seq data. tool_type: python primary_tool: SUPPA2 measurable_outcome: Execute skill workflow successfully with valid output within 15 minutes. allowed-tools: - read_file - run_shell_command

Splicing Quantification

Quantify alternative splicing events as PSI (percent spliced in) values from RNA-seq data.

Event Types

Tool Selection

SUPPA2 (transcript TPM-based)

• Input: Transcript TPM from Salmon/kallisto
• Faster, requires transcript quantification
• Better for isoform-level analysis

rMATS-turbo (BAM-based)

• Input: Aligned BAM files
• Junction read counting
• Better for novel junction discovery

SUPPA2 Workflow

import subprocess
import pandas as pd

gtf_file = 'annotation.gtf'
tpm_file = 'transcript_tpm.tsv'
output_prefix = 'events'

# Step 1: Generate splicing events from annotation
subprocess.run([
    'suppa.py', 'generateEvents',
    '-i', gtf_file,
    '-o', output_prefix,
    '-f', 'ioe',  # IOE format for PSI calculation
    '-e', 'SE', 'SS', 'MX', 'RI', 'FL'  # All event types
], check=True)

# Step 2: Calculate PSI values
for event_type in ['SE', 'A5', 'A3', 'MX', 'RI']:
    ioe_file = f'{output_prefix}_{event_type}_strict.ioe'
    subprocess.run([
        'suppa.py', 'psiPerEvent',
        '-i', ioe_file,
        '-e', tpm_file,
        '-o', f'psi_{event_type}'
    ], check=True)

# Load and examine PSI values
psi_se = pd.read_csv('psi_SE.psi', sep='\t', index_col=0)
print(f'Quantified {len(psi_se)} skipped exon events')
print(psi_se.head())

rMATS-turbo Workflow

# rMATS-turbo for BAM-based quantification
rmats.py \
    --b1 condition1_bams.txt \
    --b2 condition2_bams.txt \
    --gtf annotation.gtf \
    -t paired \
    --readLength 150 \
    --nthread 8 \
    --od output_dir \
    --tmp tmp_dir \
    --statoff  # Use for quantification only, no differential testing

import pandas as pd

# Load rMATS output
se_jc = pd.read_csv('output_dir/SE.MATS.JC.txt', sep='\t')

# Calculate average PSI across samples
# IncLevel columns contain PSI values per sample
inc_cols = [c for c in se_jc.columns if c.startswith('IncLevel')]
se_jc['mean_PSI'] = se_jc[inc_cols].mean(axis=1)

# Filter for reliable events (sufficient junction reads)
# Minimum 10-20 junction reads recommended for reliable PSI
se_jc['total_junction_reads'] = se_jc['IJC_SAMPLE_1'] + se_jc['SJC_SAMPLE_1']
reliable_events = se_jc[se_jc['total_junction_reads'] >= 20]
print(f'{len(reliable_events)} events with sufficient coverage')

Quality Thresholds

Output Interpretation

PSI values range from 0 to 1:

• PSI = 1.0: Event fully included (e.g., exon always present)
• PSI = 0.5: Equal inclusion/exclusion
• PSI = 0.0: Event fully excluded (e.g., exon always skipped)

Related Skills

• differential-splicing - Compare PSI between conditions
• rna-quantification/alignment-free-quant - Generate transcript TPM for SUPPA2
• read-alignment/star-alignment - Align reads with junction detection