LLAMA

LLAMA Help

Introductory tutorials

Tutorial video 1: Adding molecules Tutorial video 2: Decorating scaffolds Tutorial video 3: Analysing results
Tutorial 1: Adding molecules Tutorial 2: Decorating scaffolds Tutorial 3: Analysing results

LLAMA Documentation

Last updated December 06 2016 @ 10:28 UTC

  1. Overview
  2. System requirements
  3. Key concepts
    1. Lead-oriented synthesis
    2. Lead-likeness penalty
    3. Functional group filter
    4. Murcko frameworks
    5. ZINC database
    6. Interpretation of analysis
    7. Topological polar surface area (tPSA)
    8. Plane of best fit (PBF)
    9. Principal moments of inertia (PMI) plots
  4. Basic workflow
    1. Libraries
    2. Adding molecules
    3. Decorating scaffolds
    4. Analysis
  5. Advanced features
    1. Sharing libraries
    2. What is SDF upload mode?
    3. What do the 'Export as SDF' and 'Export as CSV' links do?
    4. Adding a reaction
    5. Deleting molecules
  6. Acknowledgements
    1. Citing LLAMA
    2. Support
    3. References


  1. Overview

    LLAMA (Lead Likeness And Molecular Analysis) is an open-access, web based tool which decorates and analyses molecular scaffolds. The decoration is performed using a suite of pharmaceutically relevant chemical reactions and reactants to produce a virtual library of final compounds. The lead-likeness-penalty and novelty of this virtual library are then assessed using a variety of approaches.

  2. System requirements

    LLAMA is written in HTML5 and JavaScript, and requires a modern web browser to work correctly. The following browsers are known to work correctly:

    - Internet Explorer 9 and above
    - Google Chrome v43 and above
    - Mozilla FireFox v39 and above
    - Safari v7.1.8 and above

    Other browsers such as Opera and Edge have not been tested but may still work.

  3. Key concepts

    1. Lead-oriented synthesis

      Recently, Churcher et al. have described an area of chemical space such that lead compounds residing within this space would have greater flexibility and potential for development into final clinical candidates that fit well-known drug-likeness guidelines (for example Lipinski's rule-of-five). Lead-likeness is therefore the concept of how closely a lead compound fits this lead-like space.

    2. Lead-likeness penalty

      The lead-likeness penalty is a measure of how far outside lead-like space a compound lies. For each of the key properties (heavy atom count, AlogP, number of aromatic rings and an undesirable functional group filter) the compound gains penalty points dependent on how far outside the ideal space it lies. The overall score is the sum of these individual penalties.

      Below is a series of diagrams outlining the penalty point system (numbers in the boxes represent the accrued penalty):


      Examples of the scoring system in action:

    3. Functional group filter

      The following moieties are flagged as undesirable functionality to be present within clinical candidates and therefore accrue 5 penalty points in the above system:

    4. Murcko frameworks

      To obtain the Murcko framework a molecule is stripped of all side chains to leave the ring systems and any linkers between them. A Murcko framework with alpha-attachments also includes the locations of substitutions from the rings. For more information on Murcko fragments see G. W. Bemis & M. A. Murcko, The Properties of Known Drugs. 1. Molecular Frameworks, J. Med. Chem., 1996, 39 (15), 2887-2893. For example:

    5. ZINC database

      The ZINC database is a virtual library of compounds available for purchase from commercial sources. For details on how this database was collated see J. J. Irwin, T. Sterling, M. M. Mysinger, E. S. Bolstad & R. G. Coleman, ZINC: A Free Tool to Discover Chemistry for Biology, J. Chem. Inf. Model., 2012, 52 (7), 1757-1768.

    6. Interpretation of analysis

      Interpretation of the analysis shown, including the lead-likeness penalty, is, to some extent, up to you. It is highly unlikely that all molecules derived from an individual scaffold would be lead-like: the mean lead-likeness penalty of compounds derived from a lead-like scaffold is usually less than about 3.

    7. Topological polar surface area (tPSA)

      The polar surface area of a molecule is an important molecular property when designing drugs that must cross cell membranes or the blood-brain barrier. Molecules with tPSA values in excess of around 140 Å2 are less likely to cross cell membranes. Drugs that must cross the blood-brain barrier to act on the central nervous system should have tPSA values below 90 Å2.

    8. Plane of best fit (PBF)

      The plane of best fit (PBF) is a measure of molecular three-dimensionality. Higher numbers indicate a higher degree of three-dimensionality. The number is the mean atomic distance in angstroms from a theoretical plane that passes through the molecule. The plane is configured in such a way as to minimise the PBF value.

    9. Principal moments of inertia (PMI) plots

      PMI plots represent the shape distribution of the molecules in a library. The three vertices of the triangular plot represent the extremes of molecular geometry. The top left-hand corner represents a linear molecule (diacetylene), the top right-hand corner represents a spherical molecule (adamantane) and the bottom corner represents a disc-like molecule (benzene).

      As on the lead-likeness plot, the molecules are represented by coloured data points that represent the lead-likeness penalty of the molecule. The shape of the data point represents the number of decoration reactions that the molecule has undergone. Further information about a molecule is displayed when the mouse pointer is hovered over its data point. Clicking on a data point displays even more information about the molecule.

      A current trend in medicinal chemistry is to move away from the well populated linear - disc-like axis to generate molecules with more three-dimensionality. To help users to quantify three-dimensionality we provide two metrics. The first is the mean PMI plot I1 and I2 coordinates of the selected molecules. This can be helpful when many points are grouped or overlaid, which can sometimes give a distorted impression of the distribution. The second metric is a histogram of the distribution in 20 bins. The bins are defined by the faint grey diagonal lines in the plot. The higher the bin number, the more three-dimensional the molecules in that bin.

      We have disabled the PMI calculation for molecules with two or more stereocentres with undefined geometry. This is because the calculation has little meaning for such molecular representations.

      To generate the PMI coordinates for each molecule, the system randomly generates a number of 3D conformers, minimises their energy and selects the lowest-energy conformer. The system then calculates the moments of inertia in the x, y and z axes. The PMI plot I1 coordinates are calculated by dividing inertia(x) by inertia(z). The I2 coordinates are calculated by dividing inertia(y) by inertia(z).

      The minimum energy conformer that was used to calculate the PMI I1 and I2 coordinates can be obtained by exporting the library in SDF format. The PMI I1 and I2 coordinates for all molecules are also outputted in the SDF and CSV export formats. This makes it easy to import the data into your own tools if you prefer to render your PMI plots in a "house style".

  4. Basic workflow

    1. Libraries

      Libraries are collections of molecules and reactions, and are completely independent of each other. A default library is created for you when you validate your account. You may create any number of additional libraries, to do this go to the Libraries tab and select "Create a New Library". New libraries are automatically populated with a default set of commonly-used decorating reactants. To allow more rapid decoration, it is also possible to create a library with a smaller set of building blocks (LLAMA Lite).

      There are no limitations to the number of molecules and reactions that can exist in a library. However, each library has a maximum mass limit for the molecules contained within it. This is set to 700 Daltons by default, although you can change the limit on any of your libraries if you wish. For technical reasons, the absolute maximum allowed is 3,000 Daltons.

    2. Adding molecules

      To add a molecule to your library, select "Upload a New Scaffold / Reactant" from the Scaffolds, Reactants and Products tab. You can then simply draw a scaffold or a reactant and it will be added to your library. Please note that duplicate molecules are not allowed.

    3. Decorating scaffolds

      To decorate your scaffolds, go to the Decorate Scaffolds tab. Here you can enable or disable a range of common decoration reactions by simply dragging and dropping the desired reactions between the relevant columns. Once you have decided the reactions you want enabled then simply click "Decorate".

      LLAMA will then react your scaffolds with the relevant reactant molecules in the library. You will receive a notification and an email when the process is complete, which typically takes around a minute per scaffold in the library.

    4. Analysis

      Once the decoration is complete, LLAMA then analyses the products and calculates a range of molecular properties.
      RMM: The relative molecular mass of the molecule, in Daltons.
      AlogP: An estimated calculation of the octanol/water partition coefficient. Negative values indicate that the molecule is strongly hydrophillic, and values above 4 indicate that the molecule is strongly hydrophobic.
      Heavy atoms: The number of non-hydrogen atoms within the molecule.
      Chiral centres: Chiral centres are shown on the molecule diagrams by the colour gradient of the bond approaching the centre.
      Lipinski rule-of-5 failures: Lipinski's rules give an indication of how "drug-like" a molecule is. Please see Lipinski et al. for a description and rationale of the rules.
      Lead-likeness penalty: Please see the Key concepts: Lead-likeness penalty topic for a full description of this property.

      These properties are visualised in a variety of formats in the Analysis tab. The top graph shows a plot of the products' RMM vs. AlogP. The colour of each point indicates its Lead-Likeness Penalty and, if you hover over a point, a breakdown of the penalty is displayed. Clicking on a point will display much more information about the molecule that the point represents.

  5. Advanced features

    1. Sharing libraries

      Libraries can be shared with any number of other users. This can be useful if you would like to show your decorated scaffolds to your supervisor or a colleague.
      To share a library open the libraries page and click the Edit and share icon next to the library. You can then allow other users to see the library. Granting a user 'full access' will enable them to add and delete molecules and reactions, and decorate the library. Only the library's owner can delete the library.

    2. What is SDF upload mode?

      SDF files contain molecular structures and other information. When you upload an .sdf file you can choose whether the structures will be designated as scaffolds or reactants. Both types cannot be included in the same .sdf file.

    3. What do the 'Export as SDF' and 'Export as CSV' links do?

      Clicking this link generates an SDF or CSV file containing the molecules that are currently being displayed. This is context-sensitive; if you are viewing all molecules in a library, the file will contain all of the molecules in the library. If you are viewing the details of a single molecule, the file will contain only that molecule. SDF files can be opened by a wide variety of chemical analysis software tools and workflow tools such as KNIME. CSV files can easily be opened in Microsoft Excel and contain molecular structures in Canonical SMILES format.

      When exporting in SDF format the molecular structure is that of the minimum-energy conformer that was used to generate the molecule's principal moments of inertia. It is not practical to output the 3D coordinates in CSV files, so only the canonical SMILES structure is outputted.

    4. Adding a reaction

      If you have a reaction that you wish to use for decoration of your scaffolds which is not currently in the Llama tool, you can upload a SMARTS code for the reaction under the Advanced features menu. A picture outlining the reaction can also be uploaded. Please note that it is up to yourself to check the SMARTS code is correct, for help on using SMARTS see Daylight Inc's guide to SMARTS. Hartenfeller et al., 2001 contains robust SMARTS definitions for a variety of known reactions.

    5. Deleting molecules

      Molecules can be deleted from a library by selecting the Advanced settings menu on the right sidebar. On this screen you can then select "Delete some or all molecules from the library" and then choose which molecules you wish to delete.

  6. Acknowledgements

    This service was designed at the School of Chemistry, University of Leeds by:

    Prof. Adam NelsonPrincipal Investigator
    Dr. Chris EmpsonSoftware Developer
    Dr. Philip CravenPDRA, software testing
    Zachary OwenGraphic design and software testing
    Prof. Steven MarsdenCo-Investigator
    Dr. Ian ChurcherCo-Investigator, GSK
    Dr. Richard DovestonPDRA, development of lead-likeness penalty
    Dr. Stuart WarrinerAdvisor
    Dr. Mark DowSoftware testing
    Alix HortonSoftware testing
    We thank EPSRC and GSK for funding, award reference EP/J00894X/1.

    Our reaction engine makes use of the open-source Indigo and RDKit chemoinformatics toolkits. We use ZeroMQ for message-passing between components. The d3.js library is used to create the data plots. LLAMA uses Redis for in-memory caching to improve performance.



    We thank Nathan Brown, Nicholas Firth and Greg Landrum for contributing the plane of best fit code.

    AlogP values are calculated by the VCCLab AlogPS 2.1 web service.

    1. Citing LLAMA

      When publishing any data generated by LLAMA please cite the following publication:
      I. Colomer, C. J. Empson, P. Craven, Z. Owen, R. G. Doveston, I. Churcher, S. P. Marsden, A. Nelson, A divergent synthetic approach to diverse molecular scaffolds: assessment of lead-likeness using LLAMA, an open-access computational tool, Chem. Commun., 2016, 52, 7209-7212. DOI: 10.1039/c6cc03244c

    2. Support

      If you require support or would like to report a bug please contact Dr. Chris Empson.

    3. References

      A. Nadin, C. Hattotuwagama, I. Churcher, Lead-Oriented Synthesis: A New Opportunity for Synthetic Chemistry, Angew. Chem. Int. Ed., 2012, 51, 1114-1122

      R. G. Doveston, S. P. Marsden & A. Nelson, Towards the realisation of lead-oriented synthesis, Drug Discovery Today, 2014, 19 (7), 813-819

      R. G. Doveston, P. Tosatti, M. Dow, H. Y. Li, A. J. Campbell, D. House, I. Churcher, S. P. Marsden & A. Nelson, A unified lead-oriented synthesis of over fifty molecular scaffolds, Org. Biomol. Chem., 2015, 13 (3), 859-865

      G. W. Bemis & M. A. Murcko, The Properties of Known Drugs. 1. Molecular Frameworks, J. Med. Chem., 1996, 39 (15), 2887-2893

      C. A. Lipinski, F. Lombardo, B. W. Dominy and P. J. Feeney, Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings, Adv. Drug Deliv. Rev., 2001, 46 (1-3), 3-26

      J. J. Irwin, T. Sterling, M. M. Mysinger, E. S. Bolstad and R. G. Coleman, ZINC: A Free Tool to Discover Chemistry for Biology, J. Chem. Inf. Model., 2012, 52 (7), 1757-1768

      I. V. Tetko, J. Gasteiger, R. Todeschini, A. Mauri, D. Livingstone, P. Ertl, V. A. Palyulin, E.V. Radchenko, N. S. Zefirov, A. S. Makarenko, V. Y. Tanchuk and V. V. Prokopenko, Virtual computational chemistry laboratory--design and description, J. Comput. Aided Mol. Des., 2005, 19 (6), 453-463

      M. Hartenfeller, M. Eberle, P. Meier, C. Nieto-Oberhuber, K.-H. Altmann, G. Schneider, E. Jacoby and S. Renner, A Collection of Robust Organic Synthesis Reactions for In Silico Molecule Design, J. Chem. Inf. Model., 2011, 51 (12), 3093-3098

      N. C. Firth, N. Brown and J. Blagg, Plane of Best Fit: A Novel Method to Characterize the Three-Dimensionality of Molecules, J. Chem. Inf. Model., 2012, 52 (10), 2516-2525

Back to top