alrichardbollans / mining_trait_data Goto Github PK

Improve usability of metabolite searches by simplifying processes.

Improve citation guidelines for each package

For powo searches, the installation of pykew currently needs modifying as per the commit here: RBGKew/pykew@5363379

However, the relevant functions should be made to work with the existing version of pykew

Add methods to clean_plant_occurrences to thin occurrence datasets by e.g. keeping only one record per 10 x10 arcmin grid cell to limit spatial autocorrelation (as in Pengjuan Zu et al., ‘Pollen Sterols Are Associated with Phylogeny and Environment but Not with Pollinator Guilds’, New Phytologist 230, no. 3 (May 2021): 1169–84, https://doi.org/10.1111/nph.17227.)

Using pygbif, add pipeline for downloading gbif occurrences and cleaning them.

Allowing filtering by source, and better documentation of validity of sources

Currently when empty dataframes without the 'Accepted_Name' column are passed to compile_hits an error is raised. However, these dataframes should just be ignored. Moreover, when an empty list of dataframes is passed to this method an error is raised. it would be better in this instance if an empty csv is output with the usual headings

May need to update various methods related to WCVP/WCSP and POWO following:
https://powo.science.kew.org/upcoming-changes

Many of the given feature also appear in the relatively comprehensive Dr. Duke's Phytochemical and Ethnobotanical Databases, so would be helpful to incorporate this.

i.e. those not restricted taxonomically or geographically (e.g. poison vars)

The modules are set up to import everything necessary for all the methods in the module to run. However, this means that every usage of the module requires all the libraries even if only a variable is required. For example, getting the poison data ends up requiring an install of pykew from powo_searches, which is totally unnecessary. Fix this by only importing libraries within the functions where they are required.

Some methods will end up raising an error when doing name matching in cases where no data is found e.g. when search_powo returns no hits it will try to do name matching on a nonexistent column.

Add a breakdown of sources, primarily number of hits (restricted by family)