Skip to content

MOLRAPTOR

Traceable molecular fingerprint encoding

MOLRAPTOR

Documentation for molecular fingerprint encoding from ordered, user-provided SMILES. MOLRAPTOR connects each numerical representation to its source inputs, fingerprint algorithm, effective encoding profile, and input-to-output correspondence.

CI PyPI Python versions License: LGPL v3+

Pre-stable

MOLRAPTOR is currently in alpha-stage development. Public APIs may change before 1.0.

Conceptual Model

MOLRAPTOR treats fingerprint generation as more than a matrix transformation. An encoding result is interpreted through the relationship between the molecular input, the selected algorithm, its representation semantics, the effective encoding profile, and the resulting numerical matrix.

Input Ordered SMILES source identity
Algorithm Fingerprint one algorithm per execution
Representation Binary / Count / MinHash algorithm-defined semantics
Profile Effective settings explicit encoding configuration
Result Matrix + metadata numerical representation and correspondence

The fingerprint algorithm determines the representation class. MOLRAPTOR does not expose a separate representation selector.

Morgan uses a configurable effective profile. The other nine supported algorithms use fixed effective profiles in v0.5.0.

What MOLRAPTOR Preserves

01

Input identity

Preserve the supplied strings, their order, duplicates, and the distinction between valid and invalid inputs.

02

Encoding identity

Record the selected fingerprint algorithm, representation, complete effective profile, and deterministic profile hash.

03

Numerical result

Generate a representation-dependent matrix with explicit width, shape, and dtype.

04

Correspondence

Preserve the mapping between original input positions and valid matrix rows together with MOLRAPTOR and RDKit versions.

These elements are kept separate because source identity, encoding identity, and numerical identity are not equivalent. Different input strings can produce the same numerical fingerprint, and the same input sequence can be encoded under different effective profiles.

Representation Classes

MOLRAPTOR v0.5.0 supports three numerical representation classes:

Binary Coordinates encode presence or absence according to the selected fingerprint algorithm.

Count Coordinates store multiplicities in fixed-width count vectors.

MinHash Coordinates form a MinHash sketch of the encoded shingle set rather than feature-presence bits or feature counts.

The detailed algorithm-to-representation mapping, widths, dtypes, and usage examples are documented in the Usage guide.

Source-Input Boundary

MOLRAPTOR operates on user-provided SMILES. It parses those strings with RDKit to construct the molecular graph required for fingerprint calculation.

It does not retrieve molecular records from external databases or curate, harmonize, canonicalize, or replace the supplied SMILES as an input-preprocessing step.

This source-input boundary does not constrain algorithm-internal transformations required to construct a fingerprint representation.

Where to Go Next

Page Use it for
Usage Installation, supported algorithms, command-line use, Python workflows, inputs, outputs, and failure handling.
API Reference Public Python contracts, effective profiles, representation semantics, result alignment, metadata, and hashes.
Changelog Release history and version-specific changes.

For bug reports, feature requests, and questions, use the GitHub issue tracker.

For development and contribution instructions, see CONTRIBUTING.md.