User Guide
From RoadRunnerWiki
Introduction
See About Roadrunner.
Searching for Compounds by Structure/Property
A substructure can be defined graphically using JME (Java Molecular Editor, by Peter Ertl of Novartis), or by typing in a SMILES or SMARTS. SMARTS is an extension of SMILES providing greater power of expression, both for specificity or generality. Both SMILES and SMARTS are public formats originated by Daylight CIS Inc.; refer to documentation at http://www.daylight.com/.
Searching for Compounds by Similarity
An input molecule can be entered by SMILES or graphically using JME. Similarity is computed using one of two fingerprinting methods:
- Sunset Keys
- MDL Keys
and One of four similarity measures:
- Euclidean
- Hamming
- Tanimoto
- Tversky
A fingerprint is a bitmap or binary array. Structural "keys" refer to substructures or functional groups the presence or absence of which are represented by bits on or off in the fingerprint. Sunset keys are from Sunset Molecular Discovery LLC. MDL keys are based on a published set from Elsevier MDL.
The similarity measures are differing methods of calculating a real valued similarity given two fingerprints of equal size, N bits. The Euclidean measure is simply the distance in N-dimensional space. The Tversky calculation involves two parameters, alpha and beta, and by adjusting these can thereby measure similarity as substructure and similarity as superstructure. There are many similarity measures and inherently no single best one. However, for general purposes the Tanimoto coefficient is very well accepted. For a description of these and other methods see John Bradshaw's YAMS talk.
Batch Similarity
The Batch Similarity query page composes a query based on multiple input molecules. Database molecules which meet the similarity criteria relative to any of the query molecules will be returned.
Substance Report Page
A detailed report for an individual substance. Note that a substance may be comprised of zero, one, or multiple known compounds. Conceptually and normally, a substance is something which exists in a container in reality.
Compound Report Page
Provides a detailed report for a single compound. Note that a compound is defined as a single, known compound. In Roadrunner, each compound has a unique ID number and a unique canonical smiles.
Searching for Assay Definitions
Use a substring query to search for assay definitions. In Roadrunner, an assay definition is a method of assay, whereas an assay is a specific set of experimental data.
Assay Definition Report Page
Detailed description of the assay method including experimental protocol.
Assay Upload Report Page
Details of a single data upload comprising the results of one experiment. Each experiment normally involves multiple plates, all processed using the same experimental protocol, i.e. assay definition.
Searching for Plates
Search by plate set or well volume limit. A plate set normally represents a group of compounds, a library or portion thereof.
Plate Report Page
Details of a single plate, where the tabular format represents each well with substance ID. Normally 384 well plates are used (16*24).
Registration
(Functionality currently under construction)
Registration refers to the storage of one or more new compounds in the database, in a standard form. The database issues an unique ID number and generates a canonical smiles for each new compound.
Post-HTS Analysis
Analysis of screening results can serve many purposes. Two main issues are (1) compound selection based on desirable chemical properties and (2) modelling to gain understanding and predictive capability.
Currently, the Post-HTS Analysis tools allow users to begin with any compound set defined by (1) an assay definition and (2) either active or inactive.
- Filter (by properties)
- Compounds may be filtered based on one or more of 25 properties. Normally these are used to select druglike, leadlike or probelike molecules (see ref).
- Family Selection Method
- Scaffolds - scaffolds refer to common ring systems
- Cluster by Fingerprint - fingerprints are Sunset or MDL type, clustering is by Taylor's hierarchical method (see ref).
- Cluster by MCSS - maximum common substructures are perceived based on maximum bonds or atoms.
Users, Login and Sessions
References
- Simulation Analysis of Experimental Design Strategies for Screening Random Compounds as Potential New Drugs and Agrochemicals, Taylor, R. J. Chem. Inf. Comput. Sci 1995, 35, 59-67.
- "Lead-like, drug-like or "Pub-like": how different are they?", TI Oprea, TK Allu, DC Fara, RF Rad, L Ostopovici, CG Bologa, J Comp Aid Mol Des, 21, 2007, 113-119.
