Risk evidence compilation

This workflow compiles evidence that can support risk characterization or screening. It is not a completed regulatory risk assessment. It does not define an exposure scenario, select critical studies, derive points of departure, or make a regulatory conclusion.

Discover relevant CCD lists

library(ComptoxR)
library(dplyr)

lists <- ct_chemical_list_all()
fungicide_lists <- lists |>
  filter(grepl(
    "fungic",
    paste(listName, label, shortDescription, longDescription),
    ignore.case = TRUE
  )) |>
  select(listName, label, type, chemicalCount, shortDescription)

fungicide_lists
ct_chemical_list_search_by_name("S11_SWISSPEST")

Dashboard lists can change. Use list discovery to find current candidates, but use a small explicit input vector for a reproducible example.

Resolve a fixed fungicide input

input <- tibble(
  input_id = seq_len(4),
  identifier = c("captan", "chlorothalonil", "azoxystrobin",
                 "not-a-real-chemical")
)

resolved <- ct_chemical_search_equal_bulk(input$identifier)
resolution <- input |>
  left_join(
    resolved,
    by = c("identifier" = "searchValue"),
    relationship = "many-to-many"
  ) |>
  mutate(
    match_status = case_when(
      is.na(dtxsid) ~ "unmatched",
      duplicated(input_id) | duplicated(input_id, fromLast = TRUE) ~ "ambiguous",
      TRUE ~ "matched"
    )
  )

resolution |>
  select(input_id, identifier, dtxsid, preferredName, casrn, match_status)

ids <- resolution |>
  filter(match_status == "matched") |>
  distinct(dtxsid) |>
  pull()

if (!length(ids)) stop("No unambiguous DTXSIDs were resolved.")

Keep the failed input in resolution. Excluding it without a record would make the evidence table look more complete than the input process was.

Compile CCD evidence

Use bulk wrappers when they exist.

identity <- ct_chemical_detail_search_bulk(ids)
property_experimental <- ct_chemical_property_experimental_search_bulk(ids)
property_predicted <- ct_chemical_property_predicted_search_bulk(ids)
fate <- ct_chemical_fate_search_bulk(ids)
hazard <- ct_hazard_toxval_search_bulk(ids) |>
  filter(toxvalTypeSuperCategory %in% c(
    "Point of Departure",
    "Reference Value"
  )) |>
  select(dtxsid, source, toxvalType, toxvalTypeSuperCategory,
         toxvalNumeric, toxvalUnits, speciesCommon, studyType,
         quality, qcCategory)
exposure <- ct_exposure_seem_general_search_bulk(ids)
bioactivity_detail <- ct_bioactivity_data_search_bulk(ids)

Use small iteration for endpoints without bulk wrappers.

ivive <- lapply(ids, ct_hazard_adme_ivive_search) |>
  bind_rows()

bioactivity_summary <- lapply(
  ids,
  ct_bioactivity_data_summary_search_by_dtxsid
) |>
  bind_rows()

The result now contains chemical identity, physical-chemical evidence, fate and transport, selected hazard evidence, IVIVE, SEEM exposure estimates, and bioactivity summaries. DTXSID is the join key for every CCD table.

Add optional local evidence

Run these calls only after you install the local databases.

toxval_local <- toxval_results(
  dtxsid = ids,
  qc_status = "pass_or_not_determined"
)

cas_map <- identity |>
  distinct(dtxsid, casrn)

ecotox_local <- eco_results(
  casrn = cas_map$casrn,
  eco_group = "Fish",
  endpoint = c("LC50", "EC50")
) |>
  left_join(cas_map, by = c("test_cas" = "casrn"))

ToxValDB joins directly by DTXSID. ECOTOX starts with CASRN, so use the identity table to add the DTXSID immediately. Keep unmatched CAS numbers visible.

Build an evidence-availability table

evidence <- list(
  identity = identity,
  property_experimental = property_experimental,
  property_predicted = property_predicted,
  fate = fate,
  hazard = hazard,
  ivive = ivive,
  exposure = exposure,
  bioactivity_summary = bioactivity_summary
)

available <- bind_rows(lapply(names(evidence), function(evidence_type) {
  endpoint_result <- evidence[[evidence_type]]
  if (!"dtxsid" %in% names(endpoint_result)) {
    return(tibble(
      dtxsid = character(),
      evidence_type = character(),
      available = logical()
    ))
  }

  endpoint_result |>
    filter(!is.na(dtxsid)) |>
    distinct(dtxsid) |>
    mutate(evidence_type = evidence_type, available = TRUE)
}))

availability_by_dtxsid <- tidyr::crossing(
  dtxsid = ids,
  evidence_type = names(evidence)
) |>
  left_join(available, by = c("dtxsid", "evidence_type")) |>
  mutate(available = coalesce(available, FALSE)) |>
  tidyr::pivot_wider(
    names_from = evidence_type,
    values_from = available,
    values_fill = FALSE
  )

availability <- resolution |>
  distinct(input_id, identifier, dtxsid, match_status) |>
  left_join(availability_by_dtxsid, by = "dtxsid") |>
  mutate(across(all_of(names(evidence)), ~ coalesce(.x, FALSE)))

availability

An empty endpoint result means “not retrieved from this endpoint,” not “no evidence exists.” Preserve that distinction in the screening summary.

screening_summary <- identity |>
  select(dtxsid, preferredName, casrn, molFormula, averageMass) |>
  left_join(
    exposure |>
      select(dtxsid, productionVolume, units, starts_with("probability")),
    by = "dtxsid"
  ) |>
  left_join(
    bioactivity_summary |>
      select(dtxsid, ntested, nhit, cytotoxMedianUm),
    by = "dtxsid"
  )

screening_summary

Review the full source tables before interpretation. The compact table is a screening index into the evidence, not a replacement for study review or a regulatory risk assessment.