Assesses RNA-seq data quality specifically for alternative splicing analysis. Splicing analysis is more demanding than DGE on read length, depth, library prep, alignment strategy, and annotation choice. Failures silently bias PSI estimates and inflate novel-junction false positives. QC layers include experimental design audit (library prep, read length, depth, replicates), STAR 2-pass alignment, junction saturation curves, novel-vs-known junction ratio, splice-site strength scoring (MaxEntScan, SpliceAI), strandedness, and rRNA contamination.
# Python
pip install rseqc maxentpy pysam pandas matplotlib spliceai
# CLI
conda install -c bioconda regtools fastq_screen
# Reference files
# - GENCODE basic.gtf
# - BED of canonical transcripts (for RSeQC)
# - Genome FASTA matching alignment reference
Tell your AI agent what you want to do: - "Audit my RNA-seq design for splicing analysis suitability" - "Compute junction saturation curves and check whether sequencing depth is sufficient" - "Classify junctions as known vs novel and flag samples with low known-junction fraction" - "Score 5' and 3' splice sites with MaxEntScan and SpliceAI" - "Verify library strandedness and recommend STAR 2-pass strategy"
"Will PE 100nt poly(A)-selected libraries at 30M reads/sample support splicing analysis? If not, what should change?"
"Run junction_saturation.py and junction_annotation.py across all my BAMs and report samples below acceptable thresholds."
"Configure cohort-style STAR 2-pass alignment for 12 samples - collect SJ.out.tab from pass 1 and re-align everything in pass 2 with the augmented junction set."
"For my candidate cryptic splice sites, compute MaxEntScan donor and acceptor scores plus SpliceAI in-vivo usage probability."
"I'm seeing high novel-junction rate (>40%); diagnose whether it's annotation gaps, mapping artifacts, contamination, or biology (TDP-43, SF3B1)."