[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EF6SC8V8B71QJ1JBC15F0K":3,"skill-same:01M7EF6SC8V8B71QJ1JBC15F0K":62},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":30,"commitSha":31,"skillDirectory":32,"forksCount":14,"repositoryHtmlUrl":33,"repositoryHomepageUrl":33,"ownerHtmlUrl":34,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":35,"install":36,"resources":49},"01M7EF6SC8V8B71QJ1JBC15F0K","bio-read-qc-quality-reports",null,"Generates and interprets per-file and cross-sample QC reports from FASTQ data with FastQC, falco, and MultiQC, covering Phred quality, per-base composition, GC, duplication, overrepresented sequences, and adapter content. Use when performing initial QC on raw sequencing reads, validating preprocessing, or judging a multi-sample cohort for outliers and batch effects. For long reads use NanoPlot; for adapter\u002Fquality remediation route to adapter-trimming, quality-filtering, or fastp-workflow.","Mr-Nobody-Anonymous\u002FAll-skills","Mr-Nobody-Anonymous","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184253119?v=4","01M7AEXFJHKTYQW5R0ZCQQTWFV","anonymous","01M7C00M3GQR4QKP9GZA42GFAK",0,"MIT","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-read-qc-quality-reports\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: bio-read-qc-quality-reports\ndescription: Generates and interprets per-file and cross-sample QC reports from FASTQ data with FastQC, falco, and MultiQC, covering Phred quality, per-base composition, GC, duplication, overrepresented sequences, and adapter content. Use when performing initial QC on raw sequencing reads, validating preprocessing, or judging a multi-sample cohort for outliers and batch effects. For long reads use NanoPlot; for adapter\u002Fquality remediation route to adapter-trimming, quality-filtering, or fastp-workflow.\ntool_type: cli\nprimary_tool: fastqc\nsource: \"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills\"\nsource_repository: \"GPTomics\u002FbioSkills\"\nsource_path: \"read-qc\u002Fquality-reports\u002FSKILL.md\"\nlicense: \"MIT\"\nimported_at: \"2026-09-20\"\n---\n\n## Version Compatibility\n\nReference examples tested with: FastQC 0.12+, MultiQC 1.21+, falco 1.2+, seqkit 2.5+\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- CLI: `\u003Ctool> --version` then `\u003Ctool> --help` to confirm flags\n- Python: `pip show \u003Cpackage>` then `help(module.function)` to check signatures\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed\npackage and adapt the example to match the actual API rather than retrying.\n\n# Quality Reports -- the traffic light is a hypothesis about WGS DNA, not a verdict\n\nGenerate per-file QC with FastQC\u002Ffalco and aggregate the cohort with MultiQC, then READ THE PLOTS against the assay rather than trusting pass\u002Fwarn\u002Ffail.\n\n**\"Run quality control on FASTQ files\"** -> Compute per-base quality, composition, GC, duplication, and adapter profiles per file, then aggregate across samples to find outliers.\n- CLI: `fastqc -t 8 *.fastq.gz` then `multiqc .`\n- Long reads: `NanoPlot --fastq reads.fastq.gz` (FastQC assumes fixed-length short reads)\n\nScope: this skill OWNS raw-FASTQ QC reporting and interpretation, and carries the cross-cutting quality-score \u002F chemistry \u002F duplication concepts the rest of read-qc depends on. Remediation lives elsewhere -> read-qc\u002Fadapter-trimming, read-qc\u002Fquality-filtering, read-qc\u002Ffastp-workflow. Contamination -> read-qc\u002Fcontamination-screening. Transcriptome QC on the BAM -> read-qc\u002Frnaseq-qc. OUT OF SCOPE: any modification of the reads.\n\n## The Single Most Important Modern Insight\n\n1. **FastQC pass\u002Fwarn\u002Ffail are heuristics calibrated to random whole-genome DNA, so they FALSE-FAIL on every other assay.** RNA-seq fails per-base content (random-hexamer bias) and duplication (high-expression molecules); amplicon fails duplication and GC by design; bisulfite fails base content (C->T conversion); small-RNA fails length distribution; single-cell R1 fails everything (it is barcode+UMI, not biology). A red light is a HYPOTHESIS about a WGS library. For any other protocol, first ask \"is this module expected to deviate for this chemistry?\" before treating the red as a defect. Read the plot shape; the traffic light is calibration noise.\n\n2. **On 2-color chemistry (NextSeq, NovaSeq, MiniSeq) G is the ABSENCE of signal, so poly-G tails are called at HIGH quality and the quality plot will NOT flag them.** When a cluster runs out of template, dark cycles read as a run of Gs with high confidence. Quality trimming alone does not remove them. They surface as a 3'-end RISE in G content (per-base sequence content) and a spurious high-GC spike, and they mis-map or manufacture false somatic variants if left in. The fix is a chemistry-aware poly-G trim (fastp auto-enables it from the instrument ID; cutadapt `--nextseq-trim`), not a quality cutoff. Read the per-base CONTENT plot on any 2-color run, not just the quality plot.\n\n3. **The duplication percentage is read-level and complexity-blind: it cannot tell a PCR jackpot from genuine high abundance.** Identical reads from a highly expressed transcript, a targeted amplicon, or a ChIP\u002FATAC peak are counted as duplicates even though they are independent biological molecules. Duplication % is a function of BOTH library complexity AND sequencing depth (a good library sequenced deeply shows high duplication). It is a PROMPT to reason about library complexity (preseq), never an automatic \"remove duplicates\" -- and removing duplicates in non-UMI RNA-seq is actively wrong (read-qc\u002Fumi-processing, read-qc\u002Frnaseq-qc).\n\nBonus trap: NovaSeq\u002FNextSeq emit BINNED quality scores (RTA3 uses four values: 2, 12, 23, 37), so FastQC box plots look blocky\u002Fquantized. This is the instrument's quality table, NOT bad data and NOT something to fix. The bin edges are RTA-version-specific (NovaSeq X \u002F RTA4 differs) -- never hard-code one bin set.\n\n## Tool Taxonomy\n\n| Tool | Role | Mechanism \u002F when |\n|------|------|------------------|\n| FastQC | Per-file short-read QC (HTML + zip) | Java; the module set below; duplication\u002Foverrep from the first 100k distinct reads. The de-facto standard per-file report. |\n| falco | Drop-in FastQC re-implementation (C++) | ~3x faster, lower memory, same module names and MultiQC-compatible output. Use when FastQC throughput bottlenecks a large cohort. |\n| MultiQC | Cross-sample aggregator (SCRAPER, not a re-analyzer) | Walks directories, regex-matches each tool's log\u002Freport, parses the numbers, builds one cohort report. The unit of review for multi-sample studies. |\n| seqkit stats | Instant tabular FASTA\u002FFASTQ numbers | `seqkit stats -a`: N50, Q20%, Q30%, GC%, length quartiles. For quick numbers and assembly\u002Flong-read contexts where FastQC is the wrong shape. |\n| NanoPlot \u002F NanoComp | Long-read (ONT\u002FPacBio) QC | Read-length and quality distributions, yield, N50, length-vs-quality. The correct first pass for long reads; FastQC's fixed-length assumptions break there. |\n\n## Decision Tree by Scenario\n\n| Scenario | Use | Why |\n|----------|-----|-----|\n| Per-file Illumina short-read QC | FastQC (or falco) | Module-level diagnostics; read the plots by assay |\n| Many samples \u002F a study cohort | FastQC\u002Ffalco then MultiQC | Outlier and batch detection is RELATIVE; only visible overlaid |\n| Long reads (ONT\u002FPacBio) | NanoPlot \u002F NanoComp | FastQC is built for fixed-length short reads |\n| Instant numbers, assembly input | seqkit stats -a | N50\u002FQ20\u002FQ30\u002FGC in one line; no HTML overhead |\n| Large cohort, FastQC too slow | falco then MultiQC | Same output, ~3x faster |\n\nDefault when uncertain: FastQC on each file, then MultiQC over the run directory, and judge each sample against the cohort.\n\n## FastQC Modules -- thresholds and the expert read\n\nThresholds are FastQC's `limits.txt` defaults (calibrated to random WGS DNA). The expert read is what to conclude BEYOND the traffic light.\n\n| Module | Default warn \u002F fail | Expert read |\n|--------|--------------------|-------------|\n| Per base sequence quality | warn LQ\u003C10 or median\u003C25; fail LQ\u003C5 or median\u003C20 | 3' decay is normal; blocky boxes on NovaSeq are binning; this plot will NOT reveal poly-G on 2-color |\n| Per tile sequence quality | spatial deviation (no numeric) | A hot tile band across cycles = a localized flowcell problem (bubble, debris, edge); reason no MultiQC table replaces raw FastQC |\n| Per sequence quality scores | distribution of per-read mean Q | A low-Q hump = a junk subpopulation to FILTER (not trim) |\n| Per base sequence content | warn dev>10%; fail dev>20% | First ~12 bp skew = random-hexamer priming (Hansen 2010), expected for RNA-seq, do NOT trim it. A 3'-end skew is poly-G \u002F adapter -- act on that |\n| Per sequence GC content | warn dev>15%; fail dev>30% | SHAPE matters: bimodal\u002Fsecondary peak = contamination; sharp spike = adapter dimer \u002F overrepresented; a shifted single peak = wrong-GC reference assumption |\n| Per base N content | warn N>5%; fail N>20% | Ns at a fixed position = a failed cycle; rising 3' Ns = dying clusters |\n| Sequence length distribution | warn if lengths differ; fail if any length 0 | WARNs trivially after trimming and on long reads -- ignore for those |\n| Sequence duplication levels | warn if \u003C70% would remain; fail if \u003C50% | Read-level, complexity-blind (see insight 3); high = think complexity, not dedup |\n| Overrepresented sequences | warn >0.1%; fail >1% | Most diagnostic module: it prints the sequence -- BLAST it (adapter dimer, rRNA, primer, poly-G) |\n| Adapter content | warn k-mer>5%; fail >10% | A curve climbing toward 3' = read-through from short inserts; this panel IS an insert-size readout (route to adapter-trimming) |\n| K-mer content | (deprecated, off by default) | Only appears in old reports; do not build guidance on it |\n\nAlgorithm note (why duplication\u002Foverrep are estimates): FastQC tracks only the first 100,000 DISTINCT sequences, keys on the first 50 bp for reads >75 bp (so 3' errors do not fragment a duplicate family), counts by exact identity, and extrapolates the \"% remaining if deduplicated\" headline. It is a sample-based estimate, not a full-library dedup.\n\n## Duplication Taxonomy -- four causes, four actions\n\n| Class | Mechanism | Detected by | Action |\n|-------|-----------|-------------|--------|\n| Optical | One real cluster mis-segmented (non-patterned flowcell) | Same tile, pixel distance (Picard default 100) | Removable; spatially local artifact |\n| ExAmp \u002F patterned | One molecule seeds two nanowells (HiSeq X\u002F4000, NovaSeq) | Spatially clustered, larger radius (Picard 2500 for patterned) | Removable; the reason patterned flowcells need the bigger pixel distance |\n| PCR | Same fragment amplified and sequenced twice | Identical 5' coordinates post-alignment (+UMI if present) | Mark\u002Fremove for variant calling; NEVER coordinate-dedup amplicon (use UMIs) |\n| Natural \u002F biological | Independent identical molecules (high coverage, expressed genes, amplicon start) | Indistinguishable from PCR at read level without UMIs | KEEP -- removing biases quantification (do not dedup non-UMI RNA-seq) |\n\nThe read-level duplication % FastQC reports cannot separate these. Use preseq (Daley & Smith 2013) to model the complexity curve and ask \"how many NEW molecules would more sequencing buy?\" -- that curve, not a single %, judges whether a library is exhausted or just deeply sequenced.\n\n## Quality scores and encoding\n\nPhred Q = -10*log10(P_error): Q20 = 1% error, Q30 = 0.1%, Q40 = 0.01%. Q30 is the routine Illumina target; bulk Q40+ is uncommon on legacy chemistry (phasing, signal decay) and is a tell for re-binned or synthetic data on old runs, though XLEAP-SBS (NovaSeq X, NextSeq 2000) genuinely reaches Q40+. Modern data is universally Phred+33; any Phred+64 file (Illumina 1.3-1.7) feeds 31-too-high scores to a +33-assuming tool and passes garbage silently -- convert it (`seqtk seq -Q64 -V`). A quality byte below ASCII 64 (digits\u002Fpunctuation) proves +33; detection tools sample reads to break ties.\n\n## MultiQC -- the cohort is the unit of review\n\nMultiQC does NOT re-analyze data; it scrapes tool logs\u002Freports (`search_patterns.yaml`), parses the numbers, and tabulates them per sample. Consequences: it is only as good as the files left on disk and the sample-name parsing (name collisions merge samples -- check `multiqc_data\u002Fmultiqc_sources.txt`), and it reports whatever the upstream tool wrote (a wrong reference or wrong strandedness shows as a coherent-but-wrong table, not an error). Read the General Statistics table FIRST -- outliers jump out as a column anomaly -- then overlay per-base-quality \u002F GC \u002F duplication and ask whether the low-quality set maps to one lane \u002F prep batch \u002F operator. The batch effect caught here at QC is the one not chased for a month in the DE results.\n\n```bash\n# Per-file QC, then aggregate the run\nfastqc -t 8 -o qc\u002Fraw\u002F raw_data\u002F*.fastq.gz\nmultiqc qc\u002Fraw\u002F -o qc\u002Fmultiqc\u002F -f\n\n# Compare before vs after trimming in one report\nfastqc -t 8 -o qc\u002Ftrimmed\u002F trimmed\u002F*.fastq.gz\nmultiqc qc\u002F -o qc\u002Fcompare\u002F -f          # picks up both raw\u002F and trimmed\u002F\n\n# Long reads do not go through FastQC\nNanoPlot --fastq ont_reads.fastq.gz -o qc\u002Fnanoplot\u002F\n```\n\n## Common Errors\n\n| Symptom | Cause | Solution |\n|---------|-------|----------|\n| Every RNA-seq sample fails per-base content | Random-hexamer 5' bias (Hansen 2010) | Expected; do not trim the first bases |\n| High-Q reads but a 3' G-content rise on NovaSeq | 2-color poly-G (dark cycles = G) | Chemistry-aware poly-G trim (fastp \u002F cutadapt --nextseq-trim), not -q |\n| FastQC quality boxes look quantized\u002Fblocky | NovaSeq\u002FNextSeq binned qualities (RTA3) | Expected; not a defect, do not \"fix\" |\n| MultiQC merges two samples into one row | Over-aggressive name cleaning \u002F collision | Check multiqc_sources.txt; use `--fn_as_s_name` or fix names |\n| Duplication 60%, urge to dedup RNA-seq | Read-level dup is complexity-blind | Do not dedup non-UMI RNA-seq; assess complexity (preseq) |\n| FastQC crashes \u002F huge plot on long reads | Fixed-length short-read assumptions | Use NanoPlot \u002F seqkit stats instead |\n| FastQC module missing in MultiQC | The fastqc_data.txt was not on disk \u002F wrong dir | Point MultiQC at the directory holding the zip\u002Fdata files |\n\n## References\n\nde Sena Brandine G, Smith AD. 2019. Falco: high-speed FastQC emulation for quality control of sequencing data. F1000Research 8:1874.\nEwing B, Hillier L, Wendl MC, Green P. 1998. Base-calling of automated sequencer traces using phred. I. Accuracy assessment. Genome Research 8(3):175-185.\nEwing B, Green P. 1998. Base-calling of automated sequencer traces using phred. II. Error probabilities. Genome Research 8(3):186-194.\nHansen KD, Brenner SE, Dudoit S. 2010. Biases in Illumina transcriptome sequencing caused by random hexamer priming. Nucleic Acids Research 38(12):e131.\nDaley T, Smith AD. 2013. Predicting the molecular complexity of sequencing libraries. Nature Methods 10(4):325-327.\nEwels P, Magnusson M, Lundin S, Kaller M. 2016. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics 32(19):3047-3048.\nShen W, Le S, Li Y, Hu F. 2016. SeqKit: a cross-platform and ultrafast toolkit for FASTA\u002FQ file manipulation. PLoS ONE 11(10):e0163962.\nDe Coster W, D'Hert S, Schultz DT, Cruts M, Van Broeckhoven C. 2018. NanoPack: visualizing and processing long-read sequencing data. Bioinformatics 34(15):2666-2669.\n\n## Related Skills\n\nread-qc\u002Fadapter-trimming - Remove read-through adapter flagged by the adapter-content panel\nread-qc\u002Fquality-filtering - Drop low-quality reads and trim ends\nread-qc\u002Ffastp-workflow - All-in-one QC + trim, including 2-color poly-G\nread-qc\u002Fcontamination-screening - Resolve a bimodal-GC or unexpected overrepresented-sequence signal\nread-qc\u002Frnaseq-qc - Transcriptome QC (strandedness, gene-body coverage) on the aligned BAM\nsequence-io\u002Fsequence-statistics - Programmatic per-file sequence summaries\n","## Version Compatibility\n\nReference examples tested with: FastQC 0.12+, MultiQC 1.21+, falco 1.2+, seqkit 2.5+\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- CLI: `\u003Ctool> --version` then `\u003Ctool> --help` to confirm flags\n- Python: `pip show \u003Cpackage>` then `help(module.function)` to check signatures\n\nIf code throws ImportError, AttributeError, or TypeError, introspect the installed\npackage and adapt the example to match the actual API rather than retrying.\n\n# Quality Reports -- the traffic light is a hypothesis about WGS DNA, not a verdict\n\nGenerate per-file QC with FastQC\u002Ffalco and aggregate the cohort with MultiQC, then READ THE PLOTS against the assay rather than trusting pass\u002Fwarn\u002Ffail.\n\n**\"Run quality control on FASTQ files\"** -> Compute per-base quality, composition, GC, duplication, and adapter profiles per file, then aggregate across samples to find outliers.\n- CLI: `fastqc -t 8 *.fastq.gz` then `multiqc .`\n- Long reads: `NanoPlot --fastq reads.fastq.gz` (FastQC assumes fixed-length short reads)\n\nScope: this skill OWNS raw-FASTQ QC reporting and interpretation, and carries the cross-cutting quality-score \u002F chemistry \u002F duplication concepts the rest of read-qc depends on. Remediation lives elsewhere -> read-qc\u002Fadapter-trimming, read-qc\u002Fquality-filtering, read-qc\u002Ffastp-workflow. Contamination -> read-qc\u002Fcontamination-screening. Transcriptome QC on the BAM -> read-qc\u002Frnaseq-qc. OUT OF SCOPE: any modification of the reads.\n\n## The Single Most Important Modern Insight\n\n1. **FastQC pass\u002Fwarn\u002Ffail are heuristics calibrated to random whole-genome DNA, so they FALSE-FAIL on every other assay.** RNA-seq fails per-base content (random-hexamer bias) and duplication (high-expression molecules); amplicon fails duplication and GC by design; bisulfite fails base content (C->T conversion); small-RNA fails length distribution; single-cell R1 fails everything (it is barcode+UMI, not biology). A red light is a HYPOTHESIS about a WGS library. For any other protocol, first ask \"is this module expected to deviate for this chemistry?\" before treating the red as a defect. Read the plot shape; the traffic light is calibration noise.\n\n2. **On 2-color chemistry (NextSeq, NovaSeq, MiniSeq) G is the ABSENCE of signal, so poly-G tails are called at HIGH quality and the quality plot will NOT flag them.** When a cluster runs out of template, dark cycles read as a run of Gs with high confidence. Quality trimming alone does not remove them. They surface as a 3'-end RISE in G content (per-base sequence content) and a spurious high-GC spike, and they mis-map or manufacture false somatic variants if left in. The fix is a chemistry-aware poly-G trim (fastp auto-enables it from the instrument ID; cutadapt `--nextseq-trim`), not a quality cutoff. Read the per-base CONTENT plot on any 2-color run, not just the quality plot.\n\n3. **The duplication percentage is read-level and complexity-blind: it cannot tell a PCR jackpot from genuine high abundance.** Identical reads from a highly expressed transcript, a targeted amplicon, or a ChIP\u002FATAC peak are counted as duplicates even though they are independent biological molecules. Duplication % is a function of BOTH library complexity AND sequencing depth (a good library sequenced deeply shows high duplication). It is a PROMPT to reason about library complexity (preseq), never an automatic \"remove duplicates\" -- and removing duplicates in non-UMI RNA-seq is actively wrong (read-qc\u002Fumi-processing, read-qc\u002Frnaseq-qc).\n\nBonus trap: NovaSeq\u002FNextSeq emit BINNED quality scores (RTA3 uses four values: 2, 12, 23, 37), so FastQC box plots look blocky\u002Fquantized. This is the instrument's quality table, NOT bad data and NOT something to fix. The bin edges are RTA-version-specific (NovaSeq X \u002F RTA4 differs) -- never hard-code one bin set.\n\n## Tool Taxonomy\n\n| Tool | Role | Mechanism \u002F when |\n|------|------|------------------|\n| FastQC | Per-file short-read QC (HTML + zip) | Java; the module set below; duplication\u002Foverrep from the first 100k distinct reads. The de-facto standard per-file report. |\n| falco | Drop-in FastQC re-implementation (C++) | ~3x faster, lower memory, same module names and MultiQC-compatible output. Use when FastQC throughput bottlenecks a large cohort. |\n| MultiQC | Cross-sample aggregator (SCRAPER, not a re-analyzer) | Walks directories, regex-matches each tool's log\u002Freport, parses the numbers, builds one cohort report. The unit of review for multi-sample studies. |\n| seqkit stats | Instant tabular FASTA\u002FFASTQ numbers | `seqkit stats -a`: N50, Q20%, Q30%, GC%, length quartiles. For quick numbers and assembly\u002Flong-read contexts where FastQC is the wrong shape. |\n| NanoPlot \u002F NanoComp | Long-read (ONT\u002FPacBio) QC | Read-length and quality distributions, yield, N50, length-vs-quality. The correct first pass for long reads; FastQC's fixed-length assumptions break there. |\n\n## Decision Tree by Scenario\n\n| Scenario | Use | Why |\n|----------|-----|-----|\n| Per-file Illumina short-read QC | FastQC (or falco) | Module-level diagnostics; read the plots by assay |\n| Many samples \u002F a study cohort | FastQC\u002Ffalco then MultiQC | Outlier and batch detection is RELATIVE; only visible overlaid |\n| Long reads (ONT\u002FPacBio) | NanoPlot \u002F NanoComp | FastQC is built for fixed-length short reads |\n| Instant numbers, assembly input | seqkit stats -a | N50\u002FQ20\u002FQ30\u002FGC in one line; no HTML overhead |\n| Large cohort, FastQC too slow | falco then MultiQC | Same output, ~3x faster |\n\nDefault when uncertain: FastQC on each file, then MultiQC over the run directory, and judge each sample against the cohort.\n\n## FastQC Modules -- thresholds and the expert read\n\nThresholds are FastQC's `limits.txt` defaults (calibrated to random WGS DNA). The expert read is what to conclude BEYOND the traffic light.\n\n| Module | Default warn \u002F fail | Expert read |\n|--------|--------------------|-------------|\n| Per base sequence quality | warn LQ\u003C10 or median\u003C25; fail LQ\u003C5 or median\u003C20 | 3' decay is normal; blocky boxes on NovaSeq are binning; this plot will NOT reveal poly-G on 2-color |\n| Per tile sequence quality | spatial deviation (no numeric) | A hot tile band across cycles = a localized flowcell problem (bubble, debris, edge); reason no MultiQC table replaces raw FastQC |\n| Per sequence quality scores | distribution of per-read mean Q | A low-Q hump = a junk subpopulation to FILTER (not trim) |\n| Per base sequence content | warn dev>10%; fail dev>20% | First ~12 bp skew = random-hexamer priming (Hansen 2010), expected for RNA-seq, do NOT trim it. A 3'-end skew is poly-G \u002F adapter -- act on that |\n| Per sequence GC content | warn dev>15%; fail dev>30% | SHAPE matters: bimodal\u002Fsecondary peak = contamination; sharp spike = adapter dimer \u002F overrepresented; a shifted single peak = wrong-GC reference assumption |\n| Per base N content | warn N>5%; fail N>20% | Ns at a fixed position = a failed cycle; rising 3' Ns = dying clusters |\n| Sequence length distribution | warn if lengths differ; fail if any length 0 | WARNs trivially after trimming and on long reads -- ignore for those |\n| Sequence duplication levels | warn if \u003C70% would remain; fail if \u003C50% | Read-level, complexity-blind (see insight 3); high = think complexity, not dedup |\n| Overrepresented sequences | warn >0.1%; fail >1% | Most diagnostic module: it prints the sequence -- BLAST it (adapter dimer, rRNA, primer, poly-G) |\n| Adapter content | warn k-mer>5%; fail >10% | A curve climbing toward 3' = read-through from short inserts; this panel IS an insert-size readout (route to adapter-trimming) |\n| K-mer content | (deprecated, off by default) | Only appears in old reports; do not build guidance on it |\n\nAlgorithm note (why duplication\u002Foverrep are estimates): FastQC tracks only the first 100,000 DISTINCT sequences, keys on the first 50 bp for reads >75 bp (so 3' errors do not fragment a duplicate family), counts by exact identity, and extrapolates the \"% remaining if deduplicated\" headline. It is a sample-based estimate, not a full-library dedup.\n\n## Duplication Taxonomy -- four causes, four actions\n\n| Class | Mechanism | Detected by | Action |\n|-------|-----------|-------------|--------|\n| Optical | One real cluster mis-segmented (non-patterned flowcell) | Same tile, pixel distance (Picard default 100) | Removable; spatially local artifact |\n| ExAmp \u002F patterned | One molecule seeds two nanowells (HiSeq X\u002F4000, NovaSeq) | Spatially clustered, larger radius (Picard 2500 for patterned) | Removable; the reason patterned flowcells need the bigger pixel distance |\n| PCR | Same fragment amplified and sequenced twice | Identical 5' coordinates post-alignment (+UMI if present) | Mark\u002Fremove for variant calling; NEVER coordinate-dedup amplicon (use UMIs) |\n| Natural \u002F biological | Independent identical molecules (high coverage, expressed genes, amplicon start) | Indistinguishable from PCR at read level without UMIs | KEEP -- removing biases quantification (do not dedup non-UMI RNA-seq) |\n\nThe read-level duplication % FastQC reports cannot separate these. Use preseq (Daley & Smith 2013) to model the complexity curve and ask \"how many NEW molecules would more sequencing buy?\" -- that curve, not a single %, judges whether a library is exhausted or just deeply sequenced.\n\n## Quality scores and encoding\n\nPhred Q = -10*log10(P_error): Q20 = 1% error, Q30 = 0.1%, Q40 = 0.01%. Q30 is the routine Illumina target; bulk Q40+ is uncommon on legacy chemistry (phasing, signal decay) and is a tell for re-binned or synthetic data on old runs, though XLEAP-SBS (NovaSeq X, NextSeq 2000) genuinely reaches Q40+. Modern data is universally Phred+33; any Phred+64 file (Illumina 1.3-1.7) feeds 31-too-high scores to a +33-assuming tool and passes garbage silently -- convert it (`seqtk seq -Q64 -V`). A quality byte below ASCII 64 (digits\u002Fpunctuation) proves +33; detection tools sample reads to break ties.\n\n## MultiQC -- the cohort is the unit of review\n\nMultiQC does NOT re-analyze data; it scrapes tool logs\u002Freports (`search_patterns.yaml`), parses the numbers, and tabulates them per sample. Consequences: it is only as good as the files left on disk and the sample-name parsing (name collisions merge samples -- check `multiqc_data\u002Fmultiqc_sources.txt`), and it reports whatever the upstream tool wrote (a wrong reference or wrong strandedness shows as a coherent-but-wrong table, not an error). Read the General Statistics table FIRST -- outliers jump out as a column anomaly -- then overlay per-base-quality \u002F GC \u002F duplication and ask whether the low-quality set maps to one lane \u002F prep batch \u002F operator. The batch effect caught here at QC is the one not chased for a month in the DE results.\n\n```bash\n# Per-file QC, then aggregate the run\nfastqc -t 8 -o qc\u002Fraw\u002F raw_data\u002F*.fastq.gz\nmultiqc qc\u002Fraw\u002F -o qc\u002Fmultiqc\u002F -f\n\n# Compare before vs after trimming in one report\nfastqc -t 8 -o qc\u002Ftrimmed\u002F trimmed\u002F*.fastq.gz\nmultiqc qc\u002F -o qc\u002Fcompare\u002F -f          # picks up both raw\u002F and trimmed\u002F\n\n# Long reads do not go through FastQC\nNanoPlot --fastq ont_reads.fastq.gz -o qc\u002Fnanoplot\u002F\n```\n\n## Common Errors\n\n| Symptom | Cause | Solution |\n|---------|-------|----------|\n| Every RNA-seq sample fails per-base content | Random-hexamer 5' bias (Hansen 2010) | Expected; do not trim the first bases |\n| High-Q reads but a 3' G-content rise on NovaSeq | 2-color poly-G (dark cycles = G) | Chemistry-aware poly-G trim (fastp \u002F cutadapt --nextseq-trim), not -q |\n| FastQC quality boxes look quantized\u002Fblocky | NovaSeq\u002FNextSeq binned qualities (RTA3) | Expected; not a defect, do not \"fix\" |\n| MultiQC merges two samples into one row | Over-aggressive name cleaning \u002F collision | Check multiqc_sources.txt; use `--fn_as_s_name` or fix names |\n| Duplication 60%, urge to dedup RNA-seq | Read-level dup is complexity-blind | Do not dedup non-UMI RNA-seq; assess complexity (preseq) |\n| FastQC crashes \u002F huge plot on long reads | Fixed-length short-read assumptions | Use NanoPlot \u002F seqkit stats instead |\n| FastQC module missing in MultiQC | The fastqc_data.txt was not on disk \u002F wrong dir | Point MultiQC at the directory holding the zip\u002Fdata files |\n\n## References\n\nde Sena Brandine G, Smith AD. 2019. Falco: high-speed FastQC emulation for quality control of sequencing data. F1000Research 8:1874.\nEwing B, Hillier L, Wendl MC, Green P. 1998. Base-calling of automated sequencer traces using phred. I. Accuracy assessment. Genome Research 8(3):175-185.\nEwing B, Green P. 1998. Base-calling of automated sequencer traces using phred. II. Error probabilities. Genome Research 8(3):186-194.\nHansen KD, Brenner SE, Dudoit S. 2010. Biases in Illumina transcriptome sequencing caused by random hexamer priming. Nucleic Acids Research 38(12):e131.\nDaley T, Smith AD. 2013. Predicting the molecular complexity of sequencing libraries. Nature Methods 10(4):325-327.\nEwels P, Magnusson M, Lundin S, Kaller M. 2016. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics 32(19):3047-3048.\nShen W, Le S, Li Y, Hu F. 2016. SeqKit: a cross-platform and ultrafast toolkit for FASTA\u002FQ file manipulation. PLoS ONE 11(10):e0163962.\nDe Coster W, D'Hert S, Schultz DT, Cruts M, Van Broeckhoven C. 2018. NanoPack: visualizing and processing long-read sequencing data. Bioinformatics 34(15):2666-2669.\n\n## Related Skills\n\nread-qc\u002Fadapter-trimming - Remove read-through adapter flagged by the adapter-content panel\nread-qc\u002Fquality-filtering - Drop low-quality reads and trim ends\nread-qc\u002Ffastp-workflow - All-in-one QC + trim, including 2-color poly-G\nread-qc\u002Fcontamination-screening - Resolve a bimodal-GC or unexpected overrepresented-sequence signal\nread-qc\u002Frnaseq-qc - Transcriptome QC (strandedness, gene-body coverage) on the aligned BAM\nsequence-io\u002Fsequence-statistics - Programmatic per-file sequence summaries\n",{"name":5,"source":24,"license":15,"tool_type":25,"description":7,"imported_at":26,"source_path":27,"primary_tool":28,"source_repository":29},"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills","cli","2026-09-20","read-qc\u002Fquality-reports\u002FSKILL.md","fastqc","GPTomics\u002FbioSkills","main","2d1750717747097b602acf30bd09a7cbd4b4c6ea","awesome_skills\u002Fbiology\u002Fbio-read-qc-quality-reports","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous","2026-11-09 11:45:36",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":48},[38,41,44],{"tool":39,"command":40},"npx","npx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-read-qc-quality-reports",{"tool":42,"command":43},"bunx","bunx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-read-qc-quality-reports",{"tool":45,"command":46},"pnpm","pnpm dlx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-read-qc-quality-reports","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-read-qc-quality-reports\u002FSKILL.md",true,[50,56],{"path":51,"resourceType":52,"resourceTypeDesc":53,"sizeBytes":54,"sourceHtmlUrl":55},"examples\u002Frun_qc_pipeline.sh","SCRIPT","脚本",1224,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-read-qc-quality-reports\u002Fexamples\u002Frun_qc_pipeline.sh",{"path":57,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":60,"sourceHtmlUrl":61},"usage-guide.md","REFERENCE","参考文档",3777,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-read-qc-quality-reports\u002Fusage-guide.md",[63,71,78,85,92,99],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":70},"01M7FFFZPC2025JQQHPHDX154J","dei-assessment","Assess diversity, equity, and inclusion with metrics. TRIGGERS - Use when user needs help with dei-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-assessment\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":70},"01M7FFFWN8H4WA7A6Y398QKB6J","dei-action-plan","Create DEI action plans with goals and initiatives. TRIGGERS - Use when user needs help with dei-action-plan related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-action-plan\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":70},"01M7FFFSHTM0QZRVZ2NRJTJ0PY","culture-fit-assessment","Design culture add assessment processes. TRIGGERS - Use when user needs help with culture-fit-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fculture-fit-assessment\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":70},"01M7FFFMA12RKEQA38KHJSXXPS","cowork-to-code-bridge","Use an already-installed, independently verified cowork-to-code bridge to run narrowly approved actions on the user's own macOS, Linux, or WSL2 machine through a local file queue.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcowork-to-code-bridge\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":70},"01M7FFFH1F6M8YRGW5V487GH1K","counter-offer-response","Draft counter-offer responses for retention. TRIGGERS - Use when user needs help with counter-offer-response related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcounter-offer-response\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":70},"01M7FFFDGEHV8Y7J9N40HZXW52","continuous-feedback-system","Design continuous feedback systems. TRIGGERS - Use when user needs help with continuous-feedback-system related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcontinuous-feedback-system\u002FSKILL.md",[],[]]