Formats and Tools

Chemical search

AtomLens can search the Managed Repository by more than a filename. It can compare chemical structures, which means it looks at atoms and bonds rather than just text.

Starting a chemical search

• Use Search by query structure in Local Molecule Files or Managed Repository to draw a query structure and choose either Similarity or Substructure.
• Use a molecule card's context menu to find similar structures or substructure matches for that selected molecule.
• Chemical search currently searches molecule records in the Managed Repository. Reaction records are not searched by these molecule-structure tools.

Similarity versus substructure

Similarity search asks: "Which molecules look chemically similar to my query?" AtomLens represents each molecule with a compact fingerprint and compares those fingerprints with a Tanimoto score. A high score means many structural features are shared, but it does not mean the query appears exactly inside the result.

Substructure search asks: "Does this molecule contain my query as an exact part of its structure?" If you draw a benzene ring, results must contain a matching ring. If you draw two fused rings, a molecule with two separate rings connected by a single bond is not the same substructure.

How similarity search works

1. AtomLens prepares the query fingerprint. The drawn structure or selected molecule is parsed into a molecule and converted into the same circular fingerprint representation used by the Managed Repository index.
2. The repository search snapshot is loaded. AtomLens uses the current compact search snapshot, including each indexed molecule's folded fingerprint and stored structure identifiers, instead of reparsing every repository file at the start of the search.
3. Impossible candidates are skipped quickly. The fingerprint index can rule out records whose bit counts cannot reach the requested minimum Tanimoto similarity, even under the best possible overlap.
4. Fingerprint Tanimoto scores rank the candidates. AtomLens compares the sorted query fingerprint bits with each candidate fingerprint, keeps results at or above the minimum similarity, and sorts higher similarity scores first. Equal scores fall back to stable name and identifier ordering.
5. Large searches stay bounded. When a result limit is set, each worker keeps only its strongest local results, then AtomLens merges those partial top results into the final sorted page. An unlimited search can still return every record meeting the minimum score.
6. Common regions are prepared for highlighting. AtomLens gives the MCS analysis a short shared time budget so complex structures cannot delay the results indefinitely. The violet highlight shows a common region found in that time; a larger common region may exist. This work never changes the fingerprint score, ranking, or number of matches. Results remain available even when no highlight is found.
7. The shared structure is highlighted. AtomLens converts the MCS target atom indices back to the target molecule's atom IDs, highlights bonds whose endpoints are both in that atom set, and renders the result preview with a violet MCS highlight. Opening the result carries the same violet highlight into the molecule window.

Reading similarity results

• The similarity score is a fingerprint Tanimoto value from 0.0 to 1.0. Higher values mean more shared fingerprint features.
• A high similarity result can contain extra atoms or a different arrangement around the shared core. Use the violet MCS highlight to see a shared part in the target structure.
• If AtomLens cannot build an MCS graph for a particular returned molecule, the result can still appear with its fingerprint score, but no violet highlight is shown for that row.

How substructure search works

1. AtomLens prepares the query. The drawn structure or selected molecule is parsed into a chemistry graph. A graph is simply a map of atoms connected by bonds.
2. The repository index is loaded. AtomLens uses a compact SQLite/Core Data backed search index instead of reading every full molecule file at the start of the search.
3. Fast screens remove obvious non-matches. A target must have enough atoms and bonds, the required elements, and the required ring, aromatic, and local bonding features. These checks are much cheaper than full graph matching.
4. Candidates are ranked. Structures that pass the fast screens are ordered so likely useful results are checked first. AtomLens uses fingerprint overlap and how much extra structure the target has beyond the query.
5. Reusable molecule graphs are cached. AtomLens keeps parsed molecules and SMSD Pro graph handles in memory while the repository index generation is current. Imports, deletes, reindexing, and iCloud/Core Data merges invalidate stale cached data.
6. SMSD Pro checks exact graph containment. AtomLens sends candidate batches to SMSD Pro. The default matcher uses VF2++ graph matching with neighborhood filters and bit-parallel feasibility checks. Metal can help with safe screening and prioritization on Apple GPUs, but the final exact graph match still runs on the CPU.
7. Exact hits are counted exhaustively before the visible page is shown. AtomLens screens the repository, proves exact substructure hits, and keeps checking candidates so the match count is complete. It ranks all exact hits with inexpensive exact-match data, computes short-timeout MCIS metrics for a small deterministic front window, and then ranks the enriched results again. Later hits retain the cheaper exact-match score.
8. Returned matches are highlighted. After ranking and applying the requested result limit, AtomLens loads depiction content and records matching atom and bond IDs only for the rows it returns. The result preview highlights the matched substructure in red, and opening a result carries that red highlight into the molecule window.

Reading substructure results

The MCIS / Target column reports the Maximum Common Induced Subgraph size compared with the size of the target molecule. In plain language, it says how many target atoms belong to the largest exact shared part of the two structures.

• 10/12 means 10 atoms are shared in a target with 12 atoms.
• 10/14 also shares 10 atoms, but the target has more extra atoms, so it is a broader and usually less tight match.
• AtomLens sorts substructure results by MCS score and coverage, so tighter matches normally appear above broader ones. Very broad searches can use cheaper exact-match scoring outside the full-metric front window until those hits are requested or refined.

Why some searches still take time

Exact substructure matching is a graph-isomorphism problem. That means AtomLens must prove that the query atoms and bonds can be mapped onto atoms and bonds in each target. Fast screens and Metal-assisted screening reduce the work, but hard molecules or broad queries can still require CPU-heavy exact matching.

Extract tools

AtomLens validates the source first, confirms counts, and then writes files or tables with progress feedback.

Split multi-structure SDF into multiple files

1. Choose File > Extract > Split multi-structure SDF into multiple files....
2. Select an SDF file with two or more structures.
3. Confirm the count and choose Managed Repository or a destination folder.
4. AtomLens writes one repository record or one .sdf file per structure, depending on the selected destination.

Extract multi-structure SDF attributes into CSV

1. Choose File > Extract > Extract multi-structure SDF attributes into CSV table....
2. Select a multi-structure SDF file.
3. Confirm the number of parsable structures and choose the destination .csv file.
4. AtomLens writes one quoted CSV table with all discovered SDF fields.

When a structure has multiple values for one field, AtomLens joins them with |. Invalid records are skipped and reported at the end.

Split multi-reaction SMILES into multiple files

1. Choose File > Extract > Split multi-reaction SMILES into multiple files....
2. Select a multi-line Reaction SMILES source file.
3. Confirm the number of valid reactions and choose a destination folder.
4. AtomLens writes one .rsmi file per valid reaction.

Split multi-reaction SMILES into RXN files

1. Choose File > Extract > Split multi-reaction SMILES into multiple RXN files....
2. Select a multi-line Reaction SMILES source file.
3. Confirm the count, choose a destination folder, and start extraction.
4. AtomLens converts each valid reaction entry into an individual .rxn file.

On macOS, the multi-structure SDF workflow can import structures into Managed Repository or write individual files to a folder. CSV and reaction extraction workflows write to the destination you choose.

Reaction support

AtomLens opens reaction files directly, including .rsmi, .rinchi, .rxn, .rdf, and reaction-bearing .cml.

• Reaction drawings include reactants, agents, products, plus signs, and arrow layout.
• The same pan, zoom, rotate, reset, and snapshot tools work in reaction mode.
• Reaction documents use a split layout with the drawing above and the participant list below.
• If atom-map numbers are available, map-oriented visual controls are enabled automatically.
• Reaction content is also supported in Quick Look and Spotlight metadata.

Advanced notation

AtomLens continues to support modern CXSMILES features and related advanced annotations.

• Markush and R-group definitions render as a main scaffold plus a clearly separated legend box.
• Repeat-unit and link-node notation is depicted when present.
• Imported highlights, supported query atoms, query bonds, and related annotations are preserved where possible.
• Not every advanced source feature maps to every export format. For example, Export writes standard SMILES, not CXSMILES text.

Supported input formats

.mol and .sdf input may use V2000 or V3000 syntax, and standard SMILES text files may still contain CXSMILES extensions. The dedicated document source editor can also convert supported IUPAC names into structure source before you apply and render it.

Supported export formats

Use File > Export... to export the current molecule or reaction context as chemistry files. Use File > Share for SVG, PNG, or PDF depictions of the current molecule or whole reaction.

On iPad, use the molecule or reaction window’s Share button, choose ChemDraw CDX or ChemDraw CDXML, then Save to Files.

Roundtrip support

AtomLens aims to preserve the important chemical meaning of supported source formats when you inspect and export them again.

• MDL V2000/V3000 and SDF retain more coordinates, stereochemistry, Sgroups, highlights, query features, and related annotations than earlier AtomLens versions.
• RGfile support carries R-group oriented structures when the source and export target can represent that information.
• CML support is stronger for both modern molecules and reaction-bearing content.
• CXSMILES parsing and depiction is stronger for Markush, repeat/link-node, highlight, and related layers.

© 2026 Dr. Sascha Losko - Obscure Essentials. All rights reserved.