{
  "qc_version": "0.1.0",
  "spec_id": "tenx_3p_v3",
  "assay": "10x Chromium Single Cell 3' Gene Expression",
  "platform": "illumina",
  "profile": {
    "n_pairs": 40000,
    "r1_len": {
      "min": 28,
      "max": 28,
      "modal": 28
    },
    "r2_len": {
      "min": 91,
      "max": 91,
      "modal": 91
    }
  },
  "findings": [
    {
      "check_id": "r1_length",
      "title": "R1 length matches barcode + UMI",
      "verdict": "pass",
      "value": 28.0,
      "unit": "bp",
      "threshold": "== 28",
      "affected_fraction": null,
      "severity": 0.0,
      "evidence": [
        {
          "spec_ref": "read_structure.R1",
          "note": "R1 should be exactly 28 bp (16 bp cell barcode + 12 bp UMI)"
        }
      ],
      "detail": "R1 lengths span 28\u201328 bp"
    },
    {
      "check_id": "whitelist_hit_rate",
      "title": "Cell barcodes on the 10x whitelist",
      "verdict": "pass",
      "value": 0.9489,
      "unit": "fraction",
      "threshold": ">= 0.85",
      "affected_fraction": null,
      "severity": 0.0,
      "evidence": [
        {
          "spec_ref": "whitelists.cell_barcode_3M_feb2018",
          "note": "R1[0:16] should match the 3M-february-2018 gel-bead barcode whitelist"
        }
      ],
      "detail": "94.9% of cell barcodes are on the whitelist"
    },
    {
      "check_id": "tso_at_r2_start",
      "title": "R2 reads beginning with the TSO (adapter-dimer / short insert)",
      "verdict": "fail",
      "value": 0.3258,
      "unit": "fraction",
      "threshold": "< 0.05",
      "affected_fraction": 0.3258,
      "severity": 0.6516,
      "evidence": [
        {
          "spec_ref": "read_structure.R2.readthrough_chain[tso_5prime]",
          "note": "R2 should start with cDNA; a leading TSO is the hallmark of empty/short-insert products"
        }
      ],
      "detail": "32.6% of R2 start with the TSO"
    },
    {
      "check_id": "r2_adapter_readthrough",
      "title": "R2 read-through into the Illumina adapter",
      "verdict": "warn",
      "value": 0.0545,
      "unit": "fraction",
      "threshold": "< 0.02",
      "affected_fraction": 0.0545,
      "severity": 0.109,
      "evidence": [
        {
          "spec_ref": "oligos.oligo_r1_readinto_adapter",
          "note": "the AGATCGGAAGAGC adapter stem in R2 means the insert was shorter than the read length"
        }
      ],
      "detail": "5.5% of R2 contain the adapter stem AGATCGGAAGAGC"
    },
    {
      "check_id": "r2_polyg_tail",
      "title": "R2 with a poly-G no-signal tail",
      "verdict": "fail",
      "value": 0.0983,
      "unit": "fraction",
      "threshold": "< 0.01",
      "affected_fraction": 0.098275,
      "severity": 0.2948,
      "evidence": [
        {
          "spec_ref": "platform_params.dark_base",
          "note": "a poly-G 3' tail on two-color instruments is 'no signal' \u2014 typical of empty/short fragments"
        }
      ],
      "detail": "9.8% of R2 have a poly-G tail"
    }
  ],
  "plan": {
    "ranked": [
      {
        "check_id": "tso_at_r2_start",
        "severity": "high",
        "why": "32.6% of R2 reads begin with the TSO versus a <5% threshold \u2014 the direct signature of empty/short-insert products where no cDNA was captured and the sequencer read straight through the template-switch oligo. Highest severity (0.65) and the strongest single indicator of a real prep failure."
      },
      {
        "check_id": "r2_polyg_tail",
        "severity": "medium",
        "why": "9.8% of R2 carry a poly-G 'no-signal' tail against a <1% threshold (~10x over) \u2014 on this two-color Illumina platform poly-G means the fragment ended before the read did, i.e. empty/short inserts, corroborating the TSO finding."
      },
      {
        "check_id": "r2_adapter_readthrough",
        "severity": "low",
        "why": "5.5% of R2 contain the AGATCGGAAGAGC adapter stem versus <2% \u2014 read-through into the Illumina adapter confirms inserts shorter than the 91 bp read, consistent with the same short-fragment population but at lower prevalence."
      },
      {
        "check_id": "whitelist_hit_rate",
        "severity": "none",
        "why": "94.9% of cell barcodes match the 3M-february-2018 whitelist (>=85% threshold), so barcoding and demultiplexing are healthy \u2014 the defect is not on the R1/cell-identification side."
      },
      {
        "check_id": "r1_length",
        "severity": "none",
        "why": "R1 is a uniform 28 bp (16 bp barcode + 12 bp UMI), exactly as the v3/v3.1 spec requires \u2014 read structure is correct."
      }
    ],
    "root_cause": "Adapter-dimer / short-insert contamination: a large fraction of the library consists of empty or very-short cDNA products (TSO-dominated fragments), most likely from low cDNA input or inadequate size-selection/bead cleanup that failed to remove short fragments during library prep.",
    "diagnosis": "The barcode side of the library is clean \u2014 R1 is the correct 28 bp and 94.9% of cell barcodes are on the whitelist \u2014 but the insert side is failing: 32.6% of R2 reads start with the TSO, with 9.8% showing poly-G no-signal tails and 5.5% reading through into the Illumina adapter, all pointing to the same population of empty/short-insert (adapter-dimer) products. This indicates insufficient cDNA insert reached the sequencer, typically from low input or a size-selection/cleanup step that did not remove short fragments. Repeat the SPRI-bead size selection to deplete the short products (and check cDNA input on the Bioanalyzer/TapeStation before resequencing); the useful, on-target read fraction here is substantially reduced.",
    "method": "llm"
  },
  "overall": "fail",
  "eval": {
    "n": 40000,
    "predicted_affected": 13074,
    "true_affected": 11940,
    "precision": 0.9133,
    "recall": 1.0,
    "f1": 0.9547,
    "confusion": {
      "tp": 11940,
      "fp": 1134,
      "fn": 0,
      "tn": 26926
    }
  }
}