@inflexa-ai/harness 0.35.8 → 0.36.0

package/dist/workflows/target-assessment/assemblers/literature.js

@@ -1,971 +0,0 @@
-import { coverageFromFilteredRows } from "../coverage.js";
-import { classifyTrialAttribution } from "../lib/target-identity-filter.js";
-import { HIGH_EXPRESSION_TPM_THRESHOLD } from "../lib/expression-constants.js";
-export { HIGH_EXPRESSION_TPM_THRESHOLD };
-// ── Shared helpers (pulled up so extracted functions can reference them) ──
-const SAFETY_RELEVANT_ORGANS = new Set([
-    "heart",
-    "liver",
-    "kidney",
-    "brain",
-    "bone marrow",
-    "lung",
-    "thyroid",
-    "pancreas",
-    "adrenal",
-    "gonad",
-    "ovary",
-    "testis",
-    "stomach",
-    "intestine",
-    "spleen",
-    "lymph node",
-    "retina",
-    "cochlea",
-    // Reproductive — ovary/testis/gonad already above; add remaining reproductive tissues
-    "placenta",
-    "uterus",
-    "endometrium",
-    "fallopian tube",
-    "epididymis",
-    "prostate",
-    "seminal vesicle",
-    // Bone — bone marrow already above; add cortical/trabecular bone and named bones
-    "bone",
-    "tibia",
-    "femur",
-    "vertebra",
-]);
-function isSafetyRelevant(tissue, organ) {
-    const t = tissue.toLowerCase();
-    const o = (organ ?? "").toLowerCase();
-    for (const safe of SAFETY_RELEVANT_ORGANS) {
-        if (t.includes(safe) || o.includes(safe))
-            return true;
-    }
-    return false;
-}
-function makeAssociationEvidence(a) {
-    const out = [];
-    if (a.geneticAssociationScore != null) {
-        out.push({
-            source: "opentargets:genetic_association",
-            score: a.geneticAssociationScore,
-            predicate: "associated_with",
-            is_human: true,
-            metadata: { disease_id: a.diseaseId },
-        });
-    }
-    if (a.knownDrugScore != null) {
-        out.push({
-            source: "opentargets:known_drug",
-            score: a.knownDrugScore,
-            predicate: "therapeutic_target",
-            is_human: true,
-            is_clinical: true,
-            metadata: { disease_id: a.diseaseId },
-        });
-    }
-    if (a.animalModelScore != null) {
-        out.push({
-            source: "opentargets:animal_model",
-            score: a.animalModelScore,
-            predicate: "validated_in_animal_model",
-            metadata: { disease_id: a.diseaseId },
-        });
-    }
-    if (a.somaticMutationScore != null) {
-        out.push({
-            source: "opentargets:somatic_mutation",
-            score: a.somaticMutationScore,
-            predicate: "somatic_mutation_associated",
-            is_human: true,
-            metadata: { disease_id: a.diseaseId },
-        });
-    }
-    if (a.literatureScore != null) {
-        out.push({
-            source: "opentargets:literature",
-            score: a.literatureScore,
-            predicate: "literature_co_occurrence",
-            metadata: { disease_id: a.diseaseId },
-        });
-    }
-    return out;
-}
-// ── Extracted content (lines 1544–2504, 2541–2678 from index.ts) ──────
-const SPECIES_PATTERNS = [
-    { species: "mouse", rx: /\b(mouse|mice|murine)\b/i },
-    { species: "rat", rx: /\brats?\b/i },
-    { species: "macaque", rx: /\b(macaque|cynomolgus|rhesus)\b/i },
-    { species: "dog", rx: /\b(canine|beagle)\b/i },
-    { species: "human", rx: /\b(human|patient|cohort)\b/i },
-];
-const IN_VITRO_RX = /\b(organoid|spheroid|cell line|cell-line|primary cells?|cultured|culture|in vitro|cell culture|hipsc|ipsc|3d culture|microfluidic|chip|on a chip)\b/i;
-function inferSpeciesFromTitle(title) {
-    let species = null;
-    for (const p of SPECIES_PATTERNS) {
-        if (p.rx.test(title)) {
-            species = p.species;
-            break;
-        }
-    }
-    if (!species)
-        return null;
-    if (IN_VITRO_RX.test(title)) {
-        return { species, modelSystem: "complex_in_vitro" };
-    }
-    return {
-        species,
-        modelSystem: species === "human" ? "ex_vivo_human" : "in_vivo_animal",
-    };
-}
-export function assemblePreclinicalLiterature(phase2) {
-    const pubmed = phase2.phase1.collectors.pubmedIndex;
-    if (pubmed.coverage !== "available")
-        return null;
-    const all = pubmed.data.results;
-    const speciesRows = all
-        .map((r) => {
-        const inf = inferSpeciesFromTitle(r.title);
-        if (!inf || inf.species === "human")
-            return null;
-        return {
-            pmid: r.pmid,
-            claim: r.title,
-            excerpt: undefined,
-            model_system: inf.modelSystem,
-            species: inf.species,
-        };
-    })
-        .filter((r) => r !== null);
-    const truncated = speciesRows.length > 50;
-    const rows = speciesRows.slice(0, 50);
-    if (rows.length === 0)
-        return null;
-    return {
-        rows,
-        total_claim_count: speciesRows.length,
-        truncated,
-    };
-}
-export function assembleKeyPapers(phase2, internalRefCounts) {
-    const pubmed = phase2.phase1.collectors.pubmedIndex;
-    if (pubmed.coverage !== "available" || pubmed.data.results.length === 0) {
-        return null;
-    }
-    const rows = pubmed.data.results.map((r) => {
-        const yearNum = parseInt(r.year, 10);
-        return {
-            pmid: r.pmid,
-            title: r.title,
-            internal_reference_count: internalRefCounts.get(r.pmid) ?? 0,
-            year: Number.isFinite(yearNum) ? yearNum : undefined,
-        };
-    });
-    rows.sort((a, b) => b.internal_reference_count - a.internal_reference_count);
-    return rows.slice(0, 25);
-}
-export function tallyInternalReferences(evidenceArrays) {
-    const counts = new Map();
-    for (const arr of evidenceArrays) {
-        for (const e of arr) {
-            if (e.pmid)
-                counts.set(e.pmid, (counts.get(e.pmid) ?? 0) + 1);
-        }
-    }
-    return counts;
-}
-// Loss-of-function literature signal — surface forms used in PubMed
-// titles to describe knockout, knockdown, null/deficient mice, conditional
-// alleles, CRISPR-mediated ablation, and homozygous-null notation. A title
-// matches if any one pattern hits.
-const KO_LITERATURE_PATTERNS = [
-    /\bknock[- ]?outs?\b/i,
-    // Bare "knockdown" removed — too broad; siRNA/shRNA-mediated knockdown is
-    // covered by the specific pattern below. Keeping it here admitted siRNA
-    // cancer-cell-line papers that are not germline KO evidence.
-    /\bconditional (?:ko|knock[- ]?out|deletion|ablation|allele)s?\b/i,
-    /\b(?:gene|genetic|germline|somatic) (?:deletion|ablation|disruption|inactivation)\b/i,
-    /\bnull (?:mice|mouse|allele|mutants?)\b/i,
-    /\b[A-Za-z0-9]+[- ](?:deficient|null) (?:mice|mouse)\b/i,
-    /\b(?:cre[/-]?lox(?:p)?|floxed|fl\/fl)\b/i,
-    /\bcrispr(?:[/-]cas9?)?[- ]?(?:knock[- ]?out|deletion|ablation|mediated|edited)\b/i,
-    /\b(?:loss[- ]of[- ]function|haploinsufficien(?:cy|t))\b/i,
-    /\b(?:shrna|sirna|antisense oligonucleotide)[- ](?:mediated )?(?:knock[- ]?down|silencing|depletion)\b/i,
-    /\bKO\b/, // case-sensitive: the bio abbreviation, not "ko" inside ordinary words
-    /[-‐–−]\/[-‐–−]/, // -/- or unicode-dash homozygous-null notation
-];
-// Patterns that disqualify a title despite matching the inclusion list above.
-//
-// Bare "knockdown" and RNAi reagents (siRNA/shRNA) are excluded because they
-// describe transient cell-line experiments, not germline or conditional KO models.
-// Somatic LOF patterns (mutations/LOF in the context of cancer/tumour pathology)
-// are excluded because they describe tumour-sequencing findings, not animal models.
-// "/\bcancer\b/" and "/\btumou?r\b/" are intentionally NOT in this list: a
-// legitimate paper such as "Tp53 KO mice are tumor-prone" mentions the tumour
-// phenotype of the KO model and should remain included. We rely on the somatic
-// context patterns (somatic+mutation, LOF+cancer) to catch sequencing papers.
-const KO_EXCLUDE_PATTERNS = [
-    /\bsi[- ]?rna\b/i,
-    /\bsh[- ]?rna\b/i,
-    /\bknockdown\b/i,
-    /\bsomatic\b.*\b(?:mutation|variant|loss[- ]of[- ]function)\b/i,
-    /\bloss[- ]of[- ]function\b.*\b(?:cancer|tumou?r|glioblastoma|carcinoma|sarcoma|lymphoma|leukemia|melanoma)\b/i,
-    /\b(?:cancer|tumou?r|glioblastoma|carcinoma|sarcoma|lymphoma|leukemia|melanoma)\b.*\bloss[- ]of[- ]function\b/i,
-];
-export function looksLikeKoLiterature(title) {
-    if (KO_EXCLUDE_PATTERNS.some((re) => re.test(title)))
-        return false;
-    return KO_LITERATURE_PATTERNS.some((rx) => rx.test(title));
-}
-export function assembleKoSupportingLiterature(phase2) {
-    const pubmed = phase2.phase1.collectors.pubmedIndex;
-    if (pubmed.coverage !== "available")
-        return [];
-    return pubmed.data.results
-        .filter((r) => looksLikeKoLiterature(r.title))
-        .slice(0, 10)
-        .map((r) => ({
-        source: "pubmed",
-        predicate: "ko_phenotype",
-        pmid: r.pmid,
-        excerpt: r.title,
-        metadata: { year: r.year, journal: r.journal },
-    }));
-}
-// ── Indications composite + filters ─────────────────────────────────
-const MEASUREMENT_TERMS = /\b(level|levels|concentration|count|measurement|content|index|score)\s*$/i;
-export function assembleMolecularInteractions(phase2, fanout) {
-    const ppi = phase2.phase1.collectors.stringPpi;
-    const polypharm = fanout?.perModulatorPolypharm.results ?? [];
-    const rows = [];
-    if (ppi.coverage === "available") {
-        const partners = dedupePpiPartners(ppi.data.partners);
-        for (const p of partners.slice(0, 50)) {
-            rows.push({
-                partner_id: p.partner_id,
-                partner_name: p.partner_name,
-                predicate: "binding",
-                best_score: p.combined_score,
-                source_count: p.sources.length,
-                paper_count: 0,
-                evidence: [
-                    {
-                        source: "string",
-                        predicate: "binding",
-                        score: p.string_score ?? 0,
-                        metadata: { combined_score: p.combined_score },
-                    },
-                ],
-            });
-        }
-    }
-    const byTarget = new Map();
-    for (const item of polypharm) {
-        if (item.coverage !== "available")
-            continue;
-        for (const hit of item.data.hits) {
-            const cur = byTarget.get(hit.targetChemblId);
-            const pch = hit.pchemblValue ?? 0;
-            if (!cur || pch > cur.pchembl) {
-                byTarget.set(hit.targetChemblId, {
-                    name: hit.targetName ?? hit.targetChemblId,
-                    pchembl: pch,
-                    modulators: cur?.modulators ?? new Set(),
-                });
-            }
-            byTarget.get(hit.targetChemblId).modulators.add(item.data.moleculeChemblId);
-        }
-    }
-    for (const [id, v] of byTarget) {
-        rows.push({
-            partner_id: id,
-            partner_name: v.name ?? id,
-            predicate: "drug_binding",
-            best_score: v.pchembl,
-            source_count: 1,
-            paper_count: 0,
-            evidence: [...v.modulators].map((m) => ({
-                source: "chembl:polypharm",
-                predicate: "drug_binding",
-                score: v.pchembl,
-                metadata: { partner_id: id, modulator: m },
-            })),
-        });
-    }
-    if (rows.length === 0)
-        return null;
-    return rows.slice(0, 100);
-}
-export function assembleBiomarkerPotential(phase2) {
-    const ot = phase2.phase1.collectors.opentargets;
-    const clinvar = phase2.phase1.collectors.clinvar;
-    const rows = [];
-    if (ot.coverage === "available") {
-        for (const a of ot.data.associations) {
-            if (a.knownDrugScore == null || a.knownDrugScore < 0.5)
-                continue;
-            rows.push({
-                partner_id: a.diseaseId,
-                partner_name: a.diseaseName,
-                predicate: "predictive",
-                best_score: a.knownDrugScore,
-                source_count: 1,
-                paper_count: 0,
-                evidence: [
-                    {
-                        source: "opentargets:known_drug",
-                        predicate: "predictive",
-                        score: a.knownDrugScore,
-                        metadata: { disease_id: a.diseaseId },
-                    },
-                ],
-                metrics: {},
-            });
-        }
-    }
-    if (clinvar.coverage === "available") {
-        const pathogenic = clinvar.data.variants.filter((v) => /pathogenic/i.test(v.clinicalSignificance));
-        if (pathogenic.length > 0) {
-            const conditions = new Map();
-            for (const v of pathogenic) {
-                for (const c of v.conditions) {
-                    if (!c)
-                        continue;
-                    conditions.set(c, (conditions.get(c) ?? 0) + 1);
-                }
-            }
-            for (const [condition, count] of conditions) {
-                rows.push({
-                    partner_id: null,
-                    partner_name: condition,
-                    predicate: "predictive",
-                    best_score: Math.min(1, count / 10),
-                    source_count: 1,
-                    paper_count: 0,
-                    evidence: [
-                        {
-                            source: "clinvar",
-                            predicate: "predictive",
-                            score: Math.min(1, count / 10),
-                            metadata: { variant_count: count },
-                        },
-                    ],
-                    metrics: {},
-                });
-            }
-        }
-    }
-    rows.sort((a, b) => b.best_score - a.best_score);
-    if (rows.length === 0)
-        return null;
-    return rows.slice(0, 50);
-}
-export function assembleResistanceEvidence(phase2) {
-    const pubmed = phase2.phase1.collectors.pubmedIndex;
-    if (pubmed.coverage !== "available")
-        return null;
-    const RESISTANCE_RX = /\b(resistance|resistant|sensiti[sz]ation|sensiti[sz]er|tolerance|escape mutation)\b/i;
-    const matches = pubmed.data.results.filter((r) => RESISTANCE_RX.test(r.title));
-    if (matches.length === 0)
-        return null;
-    const rows = matches.slice(0, 50).map((r) => ({
-        partner_id: null,
-        partner_name: r.title.slice(0, 80),
-        predicate: "resistance",
-        best_score: 0.5,
-        source_count: 1,
-        paper_count: 1,
-        evidence: [
-            {
-                source: "pubmed",
-                predicate: "resistance",
-                pmid: r.pmid,
-                excerpt: r.title,
-                metadata: { year: r.year, journal: r.journal },
-            },
-        ],
-    }));
-    return rows;
-}
-export function assembleCombinationEvidence(phase2) {
-    const pubmed = phase2.phase1.collectors.pubmedIndex;
-    if (pubmed.coverage !== "available")
-        return null;
-    const COMBINATION_RX = /\b(combination|combinations|synergy|synergistic|synergi[sz]e|synergi[sz]ed|synergi[sz]es)\b/i;
-    const matches = pubmed.data.results.filter((r) => COMBINATION_RX.test(r.title));
-    if (matches.length === 0)
-        return null;
-    const rows = matches.slice(0, 50).map((r) => ({
-        partner_id: null,
-        partner_name: r.title.slice(0, 80),
-        predicate: "combination_effect",
-        best_score: 0.5,
-        source_count: 1,
-        paper_count: 1,
-        evidence: [
-            {
-                source: "pubmed",
-                predicate: "combination_effect",
-                pmid: r.pmid,
-                excerpt: r.title,
-                metadata: { year: r.year, journal: r.journal },
-            },
-        ],
-    }));
-    return rows;
-}
-export function assembleEvidenceConflicts(phase2) {
-    const ot = phase2.phase1.collectors.opentargets;
-    if (ot.coverage !== "available")
-        return null;
-    const rows = [];
-    for (const a of ot.data.associations) {
-        const sources = [
-            { name: "ot_genetics", v: a.geneticAssociationScore },
-            { name: "ot_known_drug", v: a.knownDrugScore },
-            { name: "ot_animal_model", v: a.animalModelScore },
-            { name: "ot_somatic_mutation", v: a.somaticMutationScore },
-            { name: "literature", v: a.literatureScore },
-        ].filter((s) => s.v != null);
-        if (sources.length < 2)
-            continue;
-        const max = Math.max(...sources.map((s) => s.v));
-        const min = Math.min(...sources.map((s) => s.v));
-        if (max < 0.5 || min > 0.3)
-            continue;
-        const high = sources.find((s) => s.v === max).name;
-        const low = sources.find((s) => s.v === min).name;
-        rows.push({
-            evidence_item_id: `ot:${a.diseaseId}`,
-            predicate: `${high}=high`,
-            contradicting_predicates: [`${low}=low`],
-            surfaced_in_section: "indications",
-            evidence: makeAssociationEvidence(a),
-        });
-    }
-    if (rows.length === 0)
-        return null;
-    return rows.slice(0, 25);
-}
-export function assembleAdditionalEvidence(phase2) {
-    const ot = phase2.phase1.collectors.opentargets;
-    if (ot.coverage !== "available" || ot.data.safetyLiabilities.length === 0) {
-        return null;
-    }
-    const evidence = ot.data.safetyLiabilities.map((s) => ({
-        source: `opentargets:safety:${s.source}`,
-        predicate: "safety_liability",
-        excerpt: `${s.event}${s.effects ? ` (${s.effects})` : ""}`,
-        metadata: { biosamples: s.biosamples },
-    }));
-    return {
-        rows: [
-            {
-                predicate: "safety_liability",
-                evidence,
-            },
-        ],
-        score_floor: 0.3,
-    };
-}
-export function dedupePpiPartners(partners) {
-    const byId = new Map();
-    const byNameKey = (s) => s.trim().toLowerCase();
-    for (const p of partners) {
-        const id = p.proteinB;
-        const name = p.proteinB;
-        if (byId.has(id)) {
-            const cur = byId.get(id);
-            cur.string_score = Math.max(cur.string_score ?? 0, p.score);
-            cur.sources.add("string");
-            continue;
-        }
-        byId.set(id, {
-            partner_id: id,
-            partner_name: name,
-            string_score: p.score,
-            literature_score: null,
-            sources: new Set(["string"]),
-        });
-    }
-    const byName = new Map();
-    for (const v of byId.values()) {
-        const k = byNameKey(v.partner_name);
-        const existing = byName.get(k);
-        if (existing) {
-            existing.string_score = Math.max(existing.string_score ?? 0, v.string_score ?? 0);
-            for (const s of v.sources)
-                existing.sources.add(s);
-            continue;
-        }
-        byName.set(k, v);
-    }
-    return [...byName.values()].map((v) => ({
-        partner_id: v.partner_id,
-        partner_name: v.partner_name,
-        string_score: v.string_score,
-        literature_score: v.literature_score,
-        combined_score: (v.string_score ?? 0) + (v.literature_score ?? 0),
-        sources: [...v.sources],
-        has_human_evidence: true,
-        has_clinical_evidence: false,
-    }));
-}
-export function isClinicalMeasurement(name) {
-    return MEASUREMENT_TERMS.test(name.trim());
-}
-const CATEGORY_PATTERNS = [
-    { category: "safety", rx: /(safety\s+signal|adverse\s+event|toxicity|hepato|cardio|fatal|death)[a-z\s]*?(\.|$)/i },
-    { category: "efficacy", rx: /(lack\s+of\s+efficacy|did\s+not\s+meet\s+(?:primary|endpoint)|futility)[a-z\s]*?(\.|$)/i },
-    { category: "strategic", rx: /(strategic|business|pipeline|portfolio|sponsor\s+decision|commercial)[a-z\s]*?(\.|$)/i },
-    { category: "operational", rx: /(recruit|enroll|funding|covid|pandemic|site\s+closure|investigator|withdraw)[a-z\s]*?(\.|$)/i },
-];
-/**
- * Match the first known failure-cause pattern against `whyStopped` and return
- * the category plus the matched substring (truncated to 200 chars). When no
- * pattern matches, falls back to "operational" with the first 200 chars of
- * `whyStopped` itself — the schema requires the evidence excerpt to be a
- * literal substring of `why_stopped`, which holds by construction.
- */
-function classifyFailureCategory(whyStopped) {
-    for (const { category, rx } of CATEGORY_PATTERNS) {
-        const match = whyStopped.match(rx);
-        if (match) {
-            return { category, evidenceExcerpt: match[0].trim().slice(0, 200) };
-        }
-    }
-    return {
-        category: "operational",
-        evidenceExcerpt: whyStopped.slice(0, 200),
-    };
-}
-/**
- * Produce a discriminated `failure_category` block. The "safety" branch
- * uses `safety_evidence_excerpt`; all other branches use
- * `category_evidence_excerpt`.
- */
-export function buildFailureCategoryDiscriminated(whyStopped) {
-    const { category, evidenceExcerpt } = classifyFailureCategory(whyStopped);
-    if (category === "safety") {
-        return { category, safety_evidence_excerpt: evidenceExcerpt };
-    }
-    return { category, category_evidence_excerpt: evidenceExcerpt };
-}
-export function isSelfReference(canonicalId, diseaseId, symbol, diseaseName) {
-    if (canonicalId && diseaseId && canonicalId === diseaseId)
-        return true;
-    if (symbol && diseaseName && diseaseName.trim().toUpperCase() === symbol.trim().toUpperCase()) {
-        return true;
-    }
-    return false;
-}
-export const SOURCE_WEIGHTS = {
-    literature: 0.05,
-    ot_genetics: 0.2,
-    ot_known_drug: 0.25,
-    ot_animal_model: 0.1,
-    ot_somatic_mutation: 0.1,
-    clinvar: 0.15,
-};
-export function indicationCompositeScore(evidenceScore, sources, paperCount) {
-    const source_bonuses = {};
-    let bonusSum = 0;
-    for (const s of sources) {
-        const w = SOURCE_WEIGHTS[s];
-        if (w == null)
-            continue;
-        source_bonuses[s] = w;
-        bonusSum += w;
-    }
-    const paper_depth = paperCount <= 0 ? 0 : Math.min(Math.log10(1 + paperCount) * 0.05, 0.15);
-    const composite = evidenceScore + bonusSum + paper_depth;
-    return {
-        composite,
-        breakdown: { base: evidenceScore, source_bonuses, paper_depth },
-    };
-}
-// ── §3.10 helpers ────────────────────────────────────────────────────
-const RANK_LEVEL = {
-    absent: 0,
-    low: 1,
-    medium: 2,
-    high: 3,
-};
-export function deterministicTranslationalCommentary(phase2) {
-    const out = [];
-    const impc = phase2.phase1.collectors.impc;
-    if (impc.coverage === "available") {
-        if (impc.data.viability === "lethal_pre_weaning") {
-            out.push({
-                severity: "caution",
-                species: "mouse",
-                text: "Mouse KO is pre-weaning lethal — therapeutic window depends on partial inhibition, conditional KO, or staged dosing.",
-            });
-        }
-        else if (impc.data.viability === "subviable") {
-            out.push({
-                severity: "caution",
-                species: "mouse",
-                text: "Mouse KO is sub-viable — full inhibition tolerability uncertain; lead-finding should target partial occupancy or tissue-restricted exposure.",
-            });
-        }
-        if (impc.data.sexDimorphic) {
-            out.push({
-                severity: "caution",
-                species: "mouse",
-                text: "Mouse KO phenotype is sex-dimorphic — preclinical PK/PD and safety studies must be sex-stratified to avoid masking effects.",
-            });
-        }
-        if (impc.data.organSystems.length >= 3) {
-            out.push({
-                severity: "caution",
-                organ: impc.data.organSystems.slice(0, 3).join(", "),
-                species: "mouse",
-                text: `Mouse KO produces phenotypes across ${impc.data.organSystems.length} organ systems — pleiotropy raises the bar for tissue-selective therapeutic windows.`,
-            });
-        }
-    }
-    const exp = phase2.phase1.collectors.expressionMultiSpecies;
-    if (exp.coverage === "available") {
-        const human = exp.data.bySpecies.find((s) => s.species === "homo_sapiens");
-        if (human && human.tissues.length > 0) {
-            for (const otherSpecies of exp.data.bySpecies) {
-                if (otherSpecies.species === "homo_sapiens")
-                    continue;
-                const speciesLabel = otherSpecies.species.replace("_", " ");
-                for (const otherTissue of otherSpecies.tissues) {
-                    if (!isSafetyRelevant(otherTissue.tissue))
-                        continue;
-                    const humanTissue = human.tissues.find((t) => t.tissue.toLowerCase() === otherTissue.tissue.toLowerCase());
-                    if (!humanTissue)
-                        continue;
-                    const hLevel = RANK_LEVEL[humanTissue.rank] ?? 0;
-                    const oLevel = RANK_LEVEL[otherTissue.rank] ?? 0;
-                    if (Math.abs(hLevel - oLevel) >= 2) {
-                        out.push({
-                            severity: "caution",
-                            organ: otherTissue.tissue,
-                            species: speciesLabel,
-                            text: `Expression rank diverges by ≥2 steps between human (${humanTissue.rank}) and ${speciesLabel} (${otherTissue.rank}) in ${otherTissue.tissue} — translational extrapolation from this species is unreliable for safety in this organ.`,
-                        });
-                    }
-                }
-            }
-            // Cross-species concordance: human + ≥2 non-human species converge at
-            // `high` rank in the same safety-relevant tissue. This strengthens
-            // translational confidence for that organ.
-            for (const ht of human.tissues) {
-                if (ht.rank !== "high" || !isSafetyRelevant(ht.tissue))
-                    continue;
-                const concordantSpecies = [];
-                for (const other of exp.data.bySpecies) {
-                    if (other.species === "homo_sapiens")
-                        continue;
-                    const match = other.tissues.find((t) => t.tissue.toLowerCase() === ht.tissue.toLowerCase());
-                    if (match && match.rank === "high") {
-                        concordantSpecies.push(other.species.replace("_", " "));
-                    }
-                }
-                if (concordantSpecies.length >= 2) {
-                    out.push({
-                        severity: "ok",
-                        organ: ht.tissue,
-                        text: `${ht.tissue} expresses at high rank in human and ${concordantSpecies.join(", ")} — preclinical safety findings in this tissue translate with higher confidence.`,
-                    });
-                }
-            }
-        }
-        const queried = ["mus_musculus", "rattus_norvegicus", "macaca_mulatta", "canis_lupus_familiaris"];
-        for (const sp of queried) {
-            const tissues = exp.data.bySpecies.find((s) => s.species === sp)?.tissues ?? [];
-            if (tissues.length === 0 && !exp.data.notFound.includes(sp)) {
-                out.push({
-                    severity: "gap",
-                    species: sp.replace(/_/g, " "),
-                    text: `${sp.replace(/_/g, " ")} expression queried but no tissue calls returned — cross-species coverage is incomplete.`,
-                });
-            }
-        }
-    }
-    return out.slice(0, 6);
-}
-/**
- * Partition a list of trials into primary (on-target or unresolved) and
- * related (off-target via a related family receptor). Each trial's
- * `match_confidence` is derived from ChEMBL mechanism resolution via
- * `classifyTrialAttribution`. When no `drugChemblId` is present on an
- * intervention (because the upstream ctgov bundle carries only drug names),
- * resolution is skipped and the trial lands in primary with confidence "low".
- */
-export async function partitionTrialsByAttribution(trials, ctx) {
-    const primary = [];
-    const related = [];
-    for (const trial of trials) {
-        // Normalize interventions: ctgov bundle has string[], typed input may use objects.
-        const interventions = Array.isArray(trial.interventions)
-            ? trial.interventions.map((iv) => (typeof iv === "string" ? { drugChemblId: null, name: iv } : iv))
-            : [];
-        const result = await classifyTrialAttribution({
-            assessmentUniprot: ctx.assessmentUniprot,
-            familyUniprots: ctx.familyUniprots,
-            interventions,
-            conditions: trial.conditions ?? [],
-            drugTargetResolver: ctx.drugTargetResolver,
-        });
-        if (result.match_confidence === "off_target") {
-            related.push({ ...trial, match_confidence: "off_target" });
-        }
-        else {
-            primary.push({ ...trial, match_confidence: result.match_confidence });
-        }
-    }
-    return { primary, related };
-}
-/**
- * Classify why a candidate trial is considered relevant to the assessment.
- * Checks interventions against known class drug names first (strongest signal),
- * then falls back to title keyword matching, then condition match (weakest).
- *
- * Rows with match_confidence="low" and kind="condition_match" are unreliable
- * (they may be in the same broad disease area but use a completely different
- * drug class — e.g., SGLT2 inhibitors appearing in a GLP-1R dossier).
- * These rows are dropped by the caller before they reach the dossier.
- */
-export function classifyRelevanceBasis(trial, knownClassDrugNames, titleKeywordRx) {
-    // Check interventions against known class drugs (case-insensitive).
-    for (const iv of trial.interventions ?? []) {
-        const ivU = iv.toUpperCase();
-        for (const drug of knownClassDrugNames) {
-            if (ivU.includes(drug)) {
-                return { kind: "drug_in_class_match", drug_id: drug, matched_term: iv };
-            }
-        }
-    }
-    // Check title for target-specific keywords.
-    const titleMatch = titleKeywordRx.exec(trial.title);
-    if (titleMatch) {
-        return { kind: "title_keyword", matched_term: titleMatch[0] };
-    }
-    // Fall back to condition match.
-    return { kind: "condition_match", matched_term: trial.conditions?.[0] };
-}
-/**
- * Classify match confidence for a clinical trial row in the
- * `clinical_development.trials` section.
- *
- * Priority order:
- * 1. "high"   — gene symbol appears in intervention name or trial title.
- * 2. "medium" — intervention name matches a known class drug (drug-in-class).
- * 3. "medium" — gene symbol appears in a condition string.
- * 4. "low"    — no match on any signal.
- *
- * Drug-in-class matching uses case-insensitive substring comparison against
- * `knownClassDrugNames` (uppercase strings from ChEMBL modulators). This
- * promotes calcitonin-class interventions (e.g. "Calcitonin-salmon", "SMC021")
- * for a CALCR assessment even when the gene symbol itself is absent from the
- * intervention name.
- */
-export function classifyClinicalTrialConfidence(trial, geneSymbol, knownClassDrugNames) {
-    const symU = geneSymbol.toUpperCase();
-    const inInterventions = (trial.interventions ?? []).some((i) => i.toUpperCase().includes(symU));
-    const inTitle = trial.title.toUpperCase().includes(symU);
-    if (inInterventions || inTitle)
-        return "high";
-    // Drug-in-class: check each intervention name against known class drug names.
-    // Normalize hyphens to spaces before comparison so "Calcitonin-salmon" matches
-    // the ChEMBL preferred name "CALCITONIN SALMON".
-    for (const iv of trial.interventions ?? []) {
-        const ivU = iv.toUpperCase().replace(/-/g, " ");
-        for (const drug of knownClassDrugNames) {
-            if (ivU.includes(drug))
-                return "medium";
-        }
-    }
-    const inConditions = (trial.conditions ?? []).some((c) => c.toUpperCase().includes(symU));
-    if (inConditions)
-        return "medium";
-    return "low";
-}
-/** Rows the section carries; a candidate set beyond it is reported as truncated. */
-const DISCOVERY_TRIAL_CAP = 100;
-/** What emptied the discovery-trials section when nothing survived. */
-const DISCOVERY_TRIALS_FILTER = "trial already carried by clinical_development.trials, attributed to a related family receptor, matched only by a generic condition string at low confidence, or attribution not eligible for toxicology aggregation";
-export async function assembleDiscoveryTrials(phase2, symbol, attrCtx, knownClassDrugNames, attachAttribution) {
-    const ctgov = phase2.phase1.collectors.ctgov;
-    if (ctgov.coverage !== "available")
-        return { coverage: "queried_no_data", error: { message: "ClinicalTrials.gov unavailable" } };
-    const symU = symbol.toUpperCase();
-    // Build a case-insensitive title-keyword regex from the target symbol and
-    // common synonyms embedded in the assessment symbol name. For most targets
-    // the symbol itself (e.g. "GLP1R", "CALCR") is the right anchor; the regex
-    // allows optional separators so "GLP-1R" and "GLP 1R" also match.
-    const escapedSym = symbol.replace(/[-]/g, "[-\\s]?").replace(/\d+/, "\\d+");
-    const titleKeywordRx = new RegExp(`\\b${escapedSym}\\b`, "i");
-    // Exclude trials already covered by §2.5a (high-confidence by name/title match).
-    const candidates = ctgov.data.active.filter((t) => {
-        const inInterventions = t.interventions.some((i) => i.toUpperCase().includes(symU));
-        const inTitle = t.title.toUpperCase().includes(symU);
-        return !(inInterventions || inTitle);
-    });
-    const partitioned = candidates.length > 0 ? await partitionTrialsByAttribution(candidates, attrCtx) : { primary: [], related: [] };
-    // Classify each primary candidate's relevance basis and drop low-confidence
-    // condition-only matches — these are wrong-class contaminants (e.g., an
-    // SGLT2 trial picked up because it studies T2D while assessing GLP1R).
-    const primaryRows = partitioned.primary
-        .map((t) => {
-        const relevance_basis = classifyRelevanceBasis(t, knownClassDrugNames, titleKeywordRx);
-        return {
-            nct_id: t.nctId,
-            title: t.title,
-            phase: t.phase,
-            status: t.status,
-            conditions: t.conditions,
-            // Convert null → undefined to match DiscoveryTrialRowSchema optional fields.
-            start_date: t.startDate ?? undefined,
-            completion_date: t.primaryCompletionDate ?? undefined,
-            match_confidence: t.match_confidence,
-            relevance_basis,
-        };
-    })
-        .filter((row) => {
-        // Drop low-confidence rows whose only basis is a generic condition string.
-        // These fail the DiscoveryTrialRowSchema superRefine invariant and must
-        // not reach the dossier.
-        return !(row.match_confidence === "low" && row.relevance_basis.kind === "condition_match");
-    });
-    // Trials attributed to a related receptor rather than the assessed target
-    // are not discovery candidates at all, so they are not carried here.
-    const carried = primaryRows.slice(0, DISCOVERY_TRIAL_CAP);
-    const truncatedAway = primaryRows.length - carried.length;
-    const attributed = carried.map(attachAttribution);
-    const rows = attributed.filter((r) => r.eligible_for_toxicology_aggregation);
-    const section = coverageFromFilteredRows({
-        data: {
-            rows,
-            excluded_rows: attributed.filter((r) => !r.eligible_for_toxicology_aggregation),
-        },
-        retainedCount: rows.length,
-        // The cap is a carrying limit, not a judgement about a row, so it is
-        // reported as truncation rather than counted as a drop.
-        droppedCount: candidates.length - rows.length - truncatedAway,
-        filter: DISCOVERY_TRIALS_FILTER,
-        emptyReason: "every ClinicalTrials.gov record for the target is already carried by clinical_development.trials",
-    });
-    return truncatedAway > 0 && section.coverage === "available" ? { ...section, truncated: true } : section;
-}
-export function assembleEvidenceTimeline(phase2) {
-    const pubmed = phase2.phase1.collectors.pubmedIndex;
-    if (pubmed.coverage !== "available" || pubmed.data.results.length === 0) {
-        return null;
-    }
-    const histogram = {};
-    let first = null;
-    let last = null;
-    for (const r of pubmed.data.results) {
-        const y = parseInt(r.year, 10);
-        if (!Number.isFinite(y))
-            continue;
-        histogram[String(y)] = (histogram[String(y)] ?? 0) + 1;
-        if (first == null || y < first)
-            first = y;
-        if (last == null || y > last)
-            last = y;
-    }
-    const trend_labels = [];
-    if (first != null && last != null) {
-        const recent = Object.entries(histogram)
-            .filter(([y]) => parseInt(y, 10) >= last - 4)
-            .reduce((s, [, n]) => s + n, 0);
-        const earlier = Object.entries(histogram)
-            .filter(([y]) => parseInt(y, 10) >= last - 9 && parseInt(y, 10) < last - 4)
-            .reduce((s, [, n]) => s + n, 0);
-        if (recent > earlier * 1.5)
-            trend_labels.push("accelerating");
-        else if (recent < earlier * 0.66)
-            trend_labels.push("decelerating");
-        else
-            trend_labels.push("steady");
-    }
-    return {
-        histogram,
-        first_reported_year: first,
-        last_reported_year: last,
-        trend_labels,
-    };
-}
-const TIER_ORDER = ["computational", "basic_in_vitro", "complex_in_vitro", "in_vivo_animal", "ex_vivo_human", "clinical"];
-export function computeTranslationalChainSummary(tiers) {
-    const peakEntry = tiers.reduce((best, current) => (best === null || current.claim_count > best.claim_count ? current : best), null);
-    const peak_evidence_tier = peakEntry && peakEntry.claim_count > 0 ? peakEntry.tier : null;
-    const clinicalRow = tiers.find((t) => t.tier === "clinical");
-    const progression_complete = (clinicalRow?.claim_count ?? 0) >= 1;
-    // First present-tier whose immediate successor in TIER_ORDER has zero claims.
-    const presentByTier = new Map(tiers.map((t) => [t.tier, t.claim_count > 0]));
-    let weakest_progression_gap = null;
-    for (let i = 0; i < TIER_ORDER.length - 1; i++) {
-        const here = TIER_ORDER[i];
-        const next = TIER_ORDER[i + 1];
-        if (presentByTier.get(here) && !presentByTier.get(next)) {
-            weakest_progression_gap = here;
-            break;
-        }
-    }
-    return { peak_evidence_tier, progression_complete, weakest_progression_gap };
-}
-export function assembleTranslationalChain(phase2) {
-    const pubmed = phase2.phase1.collectors.pubmedIndex;
-    const ctgov = phase2.phase1.collectors.ctgov;
-    const impc = phase2.phase1.collectors.impc;
-    const exp = phase2.phase1.collectors.expressionMultiSpecies;
-    if (pubmed.coverage !== "available" && ctgov.coverage !== "available" && impc.coverage !== "available") {
-        return null;
-    }
-    const tierCounts = {
-        computational: { claims: 0, pmids: new Set() },
-        basic_in_vitro: { claims: 0, pmids: new Set() },
-        complex_in_vitro: { claims: 0, pmids: new Set() },
-        in_vivo_animal: { claims: 0, pmids: new Set() },
-        ex_vivo_human: { claims: 0, pmids: new Set() },
-        clinical: { claims: 0, pmids: new Set() },
-    };
-    if (impc.coverage === "available" && impc.data.phenotypeCount > 0) {
-        tierCounts.in_vivo_animal.claims += impc.data.phenotypeCount;
-    }
-    if (exp.coverage === "available") {
-        for (const s of exp.data.bySpecies) {
-            if (s.species !== "homo_sapiens")
-                tierCounts.in_vivo_animal.claims += s.tissues.length;
-        }
-    }
-    if (pubmed.coverage === "available") {
-        for (const r of pubmed.data.results) {
-            const inf = inferSpeciesFromTitle(r.title);
-            const tier = inf
-                ? inf.modelSystem === "ex_vivo_human"
-                    ? "ex_vivo_human"
-                    : inf.modelSystem === "complex_in_vitro"
-                        ? "complex_in_vitro"
-                        : "in_vivo_animal"
-                : "basic_in_vitro";
-            tierCounts[tier].claims += 1;
-            tierCounts[tier].pmids.add(r.pmid);
-        }
-    }
-    if (ctgov.coverage === "available") {
-        tierCounts.clinical.claims += ctgov.data.active.length + ctgov.data.failed.length;
-    }
-    const tierRows = TIER_ORDER.map((tier) => ({
-        tier,
-        claim_count: tierCounts[tier].claims,
-        paper_count: tierCounts[tier].pmids.size,
-    }));
-    const summary = computeTranslationalChainSummary(tierRows);
-    return {
-        tiers: tierRows,
-        peak_evidence_tier: summary.peak_evidence_tier,
-        progression_complete: summary.progression_complete,
-        weakest_progression_gap: summary.weakest_progression_gap,
-    };
-}