Research
We write up the methods, evaluations and observations that come out of building the database, and publish the data and code.
-
A Two-Axis Record of License and Provenance for Biomedical Datasets: Schema, Adjudication Rules, and Evidence Requirements
License and provenance information for biomedical datasets is scattered across license files, hosting-page metadata, documentation, papers and upstream sources, and these sources can disagree. A dataset card is specified that records license status (five values describing what published statements say) separately from hosting (four values, a decision derived from status, source-risk flags and a human sign-off).
FinalMdexa Research
-
Terms of Use of Public Databases and Terminologies in Drug Discovery: A Structured Survey
Pharmaceutical and biotechnology users of public drug-discovery databases and coding systems must locate usage terms that are spread over home pages, documentation pages and site-wide policies, and that take very different forms. For 26 widely used resources in seven categories, the pages expected to state terms were fetched and saved as text, and each stated permission and condition was recorded together with a verbatim quotation that a script confirms in the saved page.
FinalMdexa Research
-
A Classification of License Identifiers Observed on Biomedical Datasets
Datasets carry license identifiers of very different kinds: standard open licenses, data-specific agreements, software licenses, model licenses, platform placeholder values, and publisher-specific custom licenses. A classification table is described that assigns each identifier to one of five classes (permissive, permissive with conditions, restricted, manual review, treated as no statement) with the conditions it implies and the basis for the assignment.
FinalMdexa Research
-
Designing a Database Interface for Agents: Lessons from Live-Client Testing
An interface to a dataset database was first checked for protocol conformance with the official client library and then used by a real agent client to answer nine questions. The questions exposed two costs: an agent that could not tell a true "not in the database" from a failed search changed keywords seven times before answering, and an agent that wanted to report risks fetched evidence one dataset at a time.
FinalMdexa Engineering
-
Instructions for Visiting Agents: Design Principles for a Site-Level Agent Guide
A website that agents visit can tell them how to use it. The guide that Mdexa publishes for this purpose is written under a few rules: state facts and usage only, with no instruction that manipulates the agent; write for token economy and put the easiest mistakes first; publish the full text for humans as well; take every number from live data at build time; and keep the standard description in the machine-readable places where search and agent systems look, never in hidden page text.
FinalMdexa Engineering
-
A Controlled Vocabulary for Tagging Biomedical Datasets
Finding a dataset by what it contains, rather than by name, needs facet tags, and tags are only useful if they are few, controlled and honest about their evidence. A vocabulary of 12 facets and 99 values was drawn up, each value with a definition and each stored tag carrying a source grade (author-stated with a verbatim excerpt checked against a snapshot, inferred, human-confirmed or derived).
FinalMdexa Engineering
-
Retrieval over a Biomedical Dataset Database: Tokenization, Terminology Expansion, and Evaluation
Agents and people search a database of biomedical dataset cards with short Chinese, English and mixed-language queries, and a bad search is costly: an empty or wrong result leads an agent to repeat searches or to describe a different dataset's license. In the first run of an agent benchmark, 13 of 95 searches (13.
FinalMdexa Engineering
Nothing matches these filters.