About pKaAtlas
pKaAtlas integrates reported pKa records using separate source-row, representative-record and identity layers. The website now uses the frozen dataset described in manuscript versions 10.18 and 10.19.
From the full source archive to representative records
23,088 source rows are quarantined with recorded reasons; 1,805 eligible rows share representatives through within-source coordination. Cross-source records remain separate.
The representative layer contains 39,974 parsed structure entities, 1,612 source-specific protein-site context keys, and 10,380 records with unresolved identity. These are different units, not a combined compound count.
Frozen dataset v3-2026-10-13 · Count definitions match manuscripts 10.18 and 10.19. Records are not necessarily independent experiments. Experimental/predicted labels and grades are inherited annotations.
Full downloads, cleaning rules and verificationWhat the counts mean
The full archive includes every imported observation. Eligible rows pass the stated operational quarantine rules. Conservative coordination groups compatible rows within a source and preserves all source membership. Cross-source overlaps remain; representative counts do not imply independent measurements.
At the representative level, 137,111 records have parsed structures, 1,698 have protein-site contexts and 10,380 have unresolved identities. Distinct structures (39,974) and source-specific protein context keys (1,612) use different units from record counts.
Inherited labels and traceability
149,106 representatives inherit experimental labels; 83 inherit predicted labels. Grades A: 7,289; B: 128,786; C: 13,031; D: 83. Grades, experimental labels, DOI fields and imported natural-product annotations do not establish independently verified accuracy, original-paper traceability or biological origin.
Eleven inherited DOI-label acquisition groups include eighteen nested pKaHub sources and overlapping compilations. A source compilation's original size differs from its retained record count. Inherited DOI anomalies and unresolved permissions remain in the archived review fields; inspect individual records before relying on a source attribution. Epik-based charge-state assignment is not an experimental technique.
Years, solvents and scope
The representative layer covers 14 solvent labels. Original-publication-year annotations exist for 22,305 representatives, spanning 1889–1973; other year annotations describe compilations. Protein-site contexts are not unique-protein counts. Unresolved records remain searchable without being presented as parsed compounds.
Data use and citation
Retain the contributing source citations and version-specific upstream terms. DrugBank, ChEMBL and IUPAC-derived material can carry different terms; this site does not grant a blanket license for the integrated candidate. Known pending review annotations are preserved. Processing checks establish row preservation and membership conservation, not original experimental accuracy.
Cite the contributing sources relevant to the records used and identify pKaAtlas dataset version v3-2026-10-13, access date, scope and source-row IDs. A public article DOI has not been supplied.
Complete data, cleaning rules, checksums and legacy-count explanation · Inherited source registry · Current release manifest