{
  "evaluation_schema": "nmd-vcell-unipert-feasibility/1.0",
  "object_id": "UNIPERT:v1.0.0-database-resource:EA-20260729-20",
  "resource_release": "v1.0.0-database-resource",
  "evidence_freeze": "2026-07-25",
  "interface_build": "EA-20260729-20",
  "lifecycle": {
    "state": "PARTIAL_AUDIT_RELEASED_MODEL_EVALUATION_PENDING",
    "source_audit_executed": true,
    "identifier_preflight_executed": true,
    "encoder_executed": false,
    "embedding_result_count": 0,
    "predictive_benchmark_count": 0,
    "dmd_functional_outcome_count": 0
  },
  "source_audit": {
    "paper": {
      "title": "UniPert-G2CP bridges genetic and chemical screens from molecular representation to phenotype modeling",
      "journal": "Cell",
      "year": 2026,
      "doi": "10.1016/j.cell.2026.06.005",
      "pubmed_id": "42497867",
      "publication_state": "FORMAL_ARTICLE_VERIFIED"
    },
    "repository": {
      "url": "https://github.com/TencentAILabHealthcare/UniPert",
      "audited_commit": "2f5d46930dcbdeb92073a13e898abe6e363e679a",
      "audited_at": "2026-07-29",
      "code_licence": "GPL-3.0",
      "frozen_model_present": true,
      "public_g2cp_training_pipeline_present": false,
      "paper_split_and_full_benchmark_reproduction_present": false
    },
    "rights_boundary": {
      "article_access": "CC BY article",
      "code_use": "GPL-3.0 obligations apply to distributed derivative code",
      "patent_notice": "The article reports a Tencent patent application related to UniPert; freedom-to-operate is not established by this audit."
    }
  },
  "implementation_audit": {
    "supported_gene_inputs": [
      "gene name",
      "UniProt accession",
      "protein FASTA"
    ],
    "supported_compound_inputs": [
      "compound name",
      "SMILES",
      "CSV or dictionary"
    ],
    "anndata_perturbation_types": [
      "genetic",
      "chemical"
    ],
    "direct_condition_tokens_absent": [
      "perturbation direction",
      "dose",
      "time",
      "CRISPRi versus CRISPRa versus KO",
      "disease context",
      "cell state"
    ],
    "consequence": "UniPert can be assessed first as a frozen perturbagen identity representation; it cannot by itself define a DMD condition-response simulator."
  },
  "candidate_identifier_preflight": {
    "requested_candidate_count": 21,
    "matched_reviewed_human_uniprot_count": 21,
    "missing_candidate_count": 0,
    "missing_genes": [],
    "source": "UniProt REST API reviewed Homo sapiens primary entries",
    "queried_at": "2026-07-29",
    "encoder_executed": false,
    "interpretation": "A matched accession confirms identifier readiness only. It is not an encoder run, embedding result or predictive result.",
    "records": [
      {
        "gene": "ADAM10",
        "uniprot_accession": "O14672",
        "sequence_length": 748,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "CALR",
        "uniprot_accession": "P27797",
        "sequence_length": 417,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "CPEB1",
        "uniprot_accession": "Q9BZB8",
        "sequence_length": 566,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "DDX19B",
        "uniprot_accession": "Q9UMR2",
        "sequence_length": 479,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "DNAAF3",
        "uniprot_accession": "Q8N9W5",
        "sequence_length": 541,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "DNM1",
        "uniprot_accession": "Q05193",
        "sequence_length": 864,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "EHMT2",
        "uniprot_accession": "Q96KQ7",
        "sequence_length": 1210,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "EPS8L1",
        "uniprot_accession": "Q8TE68",
        "sequence_length": 723,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "GFOD2",
        "uniprot_accession": "Q3B7J2",
        "sequence_length": 385,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "INTS13",
        "uniprot_accession": "Q9NVM9",
        "sequence_length": 706,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "LMO2",
        "uniprot_accession": "P25791",
        "sequence_length": 158,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "MON1A",
        "uniprot_accession": "Q86VX9",
        "sequence_length": 652,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "MPHOSPH6",
        "uniprot_accession": "Q99547",
        "sequence_length": 160,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "NAGLU",
        "uniprot_accession": "P54802",
        "sequence_length": 743,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "RAC3",
        "uniprot_accession": "P60763",
        "sequence_length": 192,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "RNASEH2C",
        "uniprot_accession": "Q8TDP1",
        "sequence_length": 164,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "RNF8",
        "uniprot_accession": "O76064",
        "sequence_length": 485,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "WDR4",
        "uniprot_accession": "P57081",
        "sequence_length": 412,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "ZFP69B",
        "uniprot_accession": "Q9UJL9",
        "sequence_length": 534,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "ZNF133",
        "uniprot_accession": "P52736",
        "sequence_length": 654,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      },
      {
        "gene": "ZNF236",
        "uniprot_accession": "Q9UL36",
        "sequence_length": 1845,
        "organism": "Homo sapiens",
        "reviewed": true,
        "identifier_state": "IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN"
      }
    ]
  },
  "stages": [
    {
      "stage": "U0",
      "task": "Paper, code, licence, patent-notice and implementation-surface audit",
      "state": "EXECUTED",
      "result": "Source and implementation boundaries frozen in this contract."
    },
    {
      "stage": "U1",
      "task": "Candidate identifier and sequence-input preflight",
      "state": "PARTIAL_IDENTIFIER_PREFLIGHT_EXECUTED_ENCODER_NOT_RUN",
      "result": "21/21 candidates map to reviewed human UniProt primary accessions."
    },
    {
      "stage": "U2",
      "task": "Cause-space sanity checks and known-relation retrieval",
      "state": "REGISTERED_NOT_EXECUTED",
      "result": null
    },
    {
      "stage": "U3",
      "task": "Leakage-safe G2 incremental benchmark",
      "state": "REGISTERED_NOT_EXECUTED",
      "result": null
    },
    {
      "stage": "U4",
      "task": "Public gene-to-compound cross-domain learning curve",
      "state": "REGISTERED_NOT_EXECUTED",
      "result": null
    }
  ],
  "split_contract": {
    "primary_unit": "held-out target family or pathway",
    "compound_unit": "held-out Bemis–Murcko scaffold",
    "required_exposure_labels": [
      "known training exposure",
      "possible exposure",
      "verified unseen"
    ],
    "leakage_rules": [
      "No target-family member may appear in both training and G2 test.",
      "No compound scaffold may appear in both training and scaffold-held-out test.",
      "Hyperparameters and feature selection must be frozen without test outcomes.",
      "A result with unresolved pretraining exposure must be reported separately."
    ]
  },
  "permanent_baselines": [
    "ridge_native",
    "ESM2_only",
    "ECFP4_only",
    "nearest_neighbour_transfer",
    "random_embedding",
    "UniPert_frozen"
  ],
  "production_gate": "Adopt UniPert as a feature only if it improves a predeclared held-out task over simple baselines with uncertainty, survives exposure and leakage audit, and does not degrade worst-group performance.",
  "permitted_current_use": "Reference architecture and registered frozen-feature baseline for no-new-wet-lab evaluation.",
  "prohibited_current_use": "DMD treatment ranking, KO/overexpression simulation, dose/time response, mechanistic causality, safety or clinical inference.",
  "boundary": "This release verifies sources, implementation limits and 21-candidate identifier readiness. It publishes zero UniPert embeddings, zero UniPert predictive benchmarks and zero DMD functional outcomes."
}