Reverse Screen

Contact Atlas

Every small molecule co-crystallized with a protein in the Protein Data Bank that clears the index admission rule, searchable by the structure you draw and readable as the contacts it makes in each site it was solved in. Draw a query, or type a name.

Draw a structure, or paste a SMILES string onto the canvas. The drawing travels as a molfile as well as a SMILES string, so a hydrogen on an aromatic nitrogen is not lost on the way.

Search


Or search by name

Matches the component code, the drug name and its synonyms where ChEMBL has them, the ChEMBL id, and the chemical component dictionary's own systematic name.

Try a drug, or a scaffold: A whole drug finds its own contact maps; press Pharmacophore to broaden it to everything shaped like it. atorvastatin (Lipitor) imatinib (Gleevec) celecoxib (Celebrex)
sildenafil (Viagra) orforglipron
apixaban (Eliquis) the gefitinib scaffold

Results

Draw a structure or type a name, then search.

What Contact Atlas is

Contact Atlas is a search over every small molecule that has been co-crystallized with a protein in the Protein Data Bank and that clears the index admission rule. You draw a structure, or type a drug name, and it returns the ligands that match, each one opening on a contact map: a flat drawing of the ligand exactly as it was deposited, ringed by the amino acid residues it touches, with every interaction measured from the deposited coordinates rather than predicted.

It is the companion to Reverse Screen and asks the opposite question. Reverse Screen takes a molecule and ranks the proteins it might bind. Contact Atlas takes a molecule and finds the co-crystal ligands that resemble it, then shows what each of those actually does inside the site it was solved in.

The four ways to search

Substructure search

An exhaustive graph match over every ligand in the index, by the structure you draw or by a SMARTS pattern. It is never truncated and never sampled: if a ligand contains your query, it is in the answer. A pattern fingerprint narrows which molecules are read before the match runs, which cannot exclude a true hit, so the search is fast without being approximate. Results are ordered by size, smallest first, so the ligands your query accounts for most of come first.

Exact structure search

Matching on the InChIKey connectivity layer, so a stereoisomer of your query still finds it. The chemical component dictionary and a drawn structure disagree about stereochemistry far more often than they disagree about connectivity, which is why connectivity is the right layer.

Pharmacophore similarity search

The same predicted three-point pharmacophore fingerprint PharmCast generates from two-dimensional structure without conformer generation, and the same similarity Reverse Screen ranks with. This is the mode that finds a scaffold hop: two molecules that present the same features in three dimensions while sharing no common substructure. Neither of the other two modes can see that.

Search by name

Component code, drug name, trade name, research code, ChEMBL identifier, or the chemical component dictionary's own systematic name. Imatinib, Gleevec, CGP-57148B and CHEMBL941 all reach the same ligand.

What a contact map shows

Each figure is drawn from one deposited structure and one ligand. It carries the typed, directional interactions geometry supports, each labelled with the residue and the measured distance:

A gray envelope shows how enclosed each part of the ligand is. Nothing in a figure is predicted or docked. Where ChEMBL records a potency for that compound against that target, the figure carries it, so the measurement sits beside the geometry. A molecule deposited in ten entries has ten figures and can be read in each of its sites.

Where the data comes from

Structures and chemical components come from the Protein Data Bank. Protein identity and naming come from UniProt. Measured potency and drug names come from ChEMBL. A ligand enters the index when it is a non-polymer component between 175 and 900 daltons (below 250, sugars, amino acids, nucleotides and metal-containing components are excluded), in an entry solved by X-ray diffraction at 2.5 angstrom or better with a free R factor at or below 0.28, or by cryo-electron microscopy at 4.0 angstrom or better, with at least one polymer chain mapped to a UniProt accession. The index is rebuilt weekly against a fresh archive snapshot, and the page header states the snapshot date it is answering from.

Who it is for

Medicinal chemists asking what a scaffold does in the sites it has been crystallized in. Structure-based designers looking for the precedent behind an interaction. Anyone who has a molecule and wants the crystallographic record for it, or for the things shaped like it, without reading a hundred entries by hand. It is free to use and needs no account.

Built by Eidogen-Sertanty, Inc. Principal author Steven M. Muskal, ORCID 0000-0002-3487-270X.