TPTdb

TPTdb provides access to a curated dataset of observed biological thermal performance measurements across the tree of life. Version 0.2.0 contains 1,300 curves for 1,069 binomial names and 38 response types. It also exposes sampling-resolution, curve-coverage, observation-flag, source, and model-fit information so users can decide which records are appropriate for a particular analysis.

The historical package name is retained for continuity. In the manuscript and documentation, TPTdb is described as a dataset rather than as a physiological-tolerance database.

Important interpretation limits

• T_lower_observed_C and T_upper_observed_C are the lowest and highest sampled temperatures. They are not CTmin, CTmax, or physiological tolerance limits.
• T_peak_observed_C is the median of the sampled temperatures sharing the largest observed response. With an even number of tied maxima, this median can fall between two sampled temperatures. It is not a fitted optimum.
• thermal_limit_proxies.csv adds optional observed 0%, <=1%, and <=5% performance thresholds on each sampled limb. These are named performance-limit proxies and are not measured CTmin or CTmax.
• flexTPC roots are five-parameter, model-derived extrapolations. Use them only when flex_fit_status is fit_available and inspect flex_root_identifiability; even a high response-scale R-squared does not make a root a measured physiological endpoint.
• Curves with only 3-4 unique temperatures are retained and explicitly flagged as low-resolution.
• A peak at a sampled boundary indicates that one limb of the curve was not observed.
• Tukey flags identify statistical extremeness within a curve. Flagged observations are retained and are not presumed erroneous.
• Experimental protocols were not harmonized. Consult the source reference before comparing records across studies.
• The Arroyo et al. model and flexTPC are optional descriptive visualization overlays. Neither changes the observed descriptors. flexTPC was fitted only to prespecified, non-negative, two-sided records with at least seven unique temperatures.

Installation

For a downloaded source folder:

install.packages("devtools")
devtools::install("path/to/TPTdb", upgrade = "never")

upgrade = "never" is intentional. Installing TPTdb does not require replacing working copies of every dependency in your R library. In particular, do not choose All if an installation prompt offers to upgrade unrelated packages inside an active RStudio session.

From GitHub after the revised release is published:

install.packages("remotes")
remotes::install_github("amahury/TPTdb", upgrade = "never")

Usage

library(TPTdb)

tpt <- TPTdb_data("all")
head(tpt$taxonomy)
head(tpt$quality)

TPTdb_explorer()

Clean-session check

When developing the package in RStudio, do not run or source R/explorer.R before calling package-development commands. Sourcing the file creates global objects named TPTdb_data and TPTdb_explorer, which can mask the functions loaded by devtools::load_all() and produce a conflict warning even when the package itself is valid.

For a clean verification, restart the R session and run only:

devtools::document()
devtools::check()
devtools::install(upgrade = "never")
library(TPTdb)
TPTdb_explorer()

The explorer uses Shiny's built-in table renderer and does not depend on the DT package. This intentionally keeps the interactive app independent of a nonessential JavaScript-table dependency.

If a previous DT.rdb is corrupt error was seen

That message refers to the installed DT package in the user's R library, not to a database file inside TPTdb. TPTdb 0.2.0 no longer needs DT, so first restart RStudio and verify this release using the clean-session commands above.

If you also use DT for other R work and want to repair that separate package, close every R/RStudio process, open a fresh R session, and reinstall the Windows binary without upgrading the rest of the library:

remove.packages("DT")
install.packages("DT", type = "binary")

Restart R once more before loading packages. This repair is optional for TPTdb.

If restarting is inconvenient, remove any sourced copies first with rm(list = intersect(c("TPTdb_data", "TPTdb_explorer"), ls(.GlobalEnv))).

Data files

The canonical machine-readable files are in inst/extdata/:

• taxonomy.csv: curve IDs, source IDs, taxonomic fields, names, and source references.
• temperature.csv: wide-format sampled temperatures in degrees Celsius.
• response.csv: wide-format responses plus variable, unit, and legacy curve-shape labels.
• traits.csv: strictly observed temperature descriptors and observed performance maximum.
• quality.csv: sampling resolution, peak position, endpoint coverage, Tukey-flag counts, protocol status, and model-fit diagnostics.
• thermal_limit_proxies.csv: optional observed 0%, <=1%, and <=5% performance-limit proxies, five-parameter flexTPC fits, extrapolation distances, model diagnostics, and root-identifiability flags.
• observation_flags.csv: one row per temperature-response pair with within-curve Tukey flags; no flagged value is deleted.
• source_manifest.csv: unique source references and the number of curves linked to each source.
• input_compilations.csv: principal third-party input compilations, access locations, and reuse basis.
• TPTdb_full.xlsx: a human-readable workbook containing the same nine tables.

See DATA_DICTIONARY.md, THERMAL_LIMIT_PROXY_METHODS.md, DATA_CORRECTIONS.md, SOURCE_PROVENANCE_AND_REUSE.md, and TAXONOMY_MAINTENANCE.md for additional details.

Licenses

The dataset is distributed under CC BY 4.0; see LICENSE-DATA.md. The R package code is distributed under the MIT License; see LICENSE.

Citation

Please cite the versioned Zenodo record:

López-Díaz, A. J. et al. A dataset of biological thermal performance observations across the tree of life. Zenodo. https://doi.org/10.5281/zenodo.20927688 (2026).