bugs

Limitations of PaperBLAST that cause gene-article associations to be
missed:

Links from EuropePMC's text mining effort are assumed to be for open
access articles. If the article is not open access (i.e. in the
EuropePMC downloads), then (1) PaperBLAST does not even try to extract
a snippet, and (2) the link is suppressed because there should always
be a snippet for an OA link. Either the PMC links should be "blessed"
so that they are not suppressed later on, and/or the open access
status of these articles should be corrected (i.e., by fetching their
metadata from EuropePMC rather than pubmed).

Limitations of snippet identification. For example, "sp|P12273|" or
"sp|P12273" or "P10275.2" may be references to UniProt P10275, but are
missed by PaperBLAST's snippet selection. This can cause valid links to
OA articles to be suppressed.

Changes to genus names -- PaperBLAST's uses "genus_name AND
identifier" to search for references to genes. If the genus name
changes, and the sequence database (MicrobesOnline or RefSeq) and the
article choose different genus names, the article will be
missed. Possible solutions include making note of changes to genus
name (is this information available from NCBI taxonomy?) and sending
additional queries to EuropePMC, or perhaps searching for these cases
in the downloaded articles.

Hits in downloaded articles -- sometimes PaperBLAST has access to the
full text of an article but EuropePMC does not (i.e., if the article
is linked to a gene via UniProt or GeneRIF and is downloaded by
pubMunch, but has no pmcId because it is too old). In these cases, we
could search the article for additional gene identifiers.