apical | An API design and testing framework

As commented, your earlier search returns an empty data frame since the searched term is part of the larger strings in the character column, GO_NAME. Therefore, instead of == or%in% operators which expects whole word matches, consider grep to search string patterns within larger string:

I dont know a trivial way to achieve this although there should be implementations for this usecase, I dont know them. Two solution approaches:

1) You could try to use existing fuzzy search libraries/solutions like elastic search (available for i.e. MongoDB), store your searchable fields in a search string and try to teach fuzzy search how to split the parts as you need them. Probably feasible if you want fuzzy matching in your query too and dont care too much about exact results.

2) Implement your own matching. Start by implementing a similarity measure and iteratively calculate that measure over the dataset. A very simple measure would be to simply add one to the similarity score for every matching field. Assign the score to the document and sort by score after iterating over the whole collection. Now you have a sorted collection by relevance to your query.

I dont see why it should make any difference whether you match top to bottom vs bottom to top.