Help

About DDMuffin

DDMuffin is a user friendly web server that implements a deep learning approach for predicting protein-ligand affinity and the effects of missense mutations on it.


Our method combines protein and small-molecule language model embeddings with a graph convolutional network to identify key residues and ligand atoms, capturing essential interaction details.

Running predictions for protein-ligand affinity

DDMuffin Affinity Submission Page

The submission page for affinity prediction can be accessed via the menu item Run (1) on the top bar menu.

  • The Affinity Prediction (2) option allows the user to submit one protein-ligand complex for prediction.
    • Users are required to provide a complex structure (3) by uploading a file in PDB format or providing a PDB accession code.
    • The ligand 3-letter code in the PDB should be provided (4).
    • If an email address is provided (5), a link to the results page will be sent to the useras soon as DDMuffin finishes processing the submission.
    • An example of the results page is also provided (6).

Results for affinity prediction

DDMuffin Affinity Prediction page 1 DDMuffin Affinity Prediction page 2 DDMuffin Affinity Prediction page 3 DDMuffin Affinity Prediction page 4

The results page for affinity prediction will be separated into 3 different panels once computations are completed.

  • On the first panel, predicted affinity (pKi) is displayed (1) alongside with the Ligand details (2).
  • The second panel will show and interactive 3D viewer (3), which allows user to analyse the interatomic interactions between the protein and the ligand. Interactions can be hidden or displayed using the controls provided (4). The viewer can also be manipulated using buttons at the bottom of the panel (5).
  • On the third panel, DDMuffin allows users to analyse the interactions in a 2D plot (6). Here, a set of switches are also provided (7) to allow customisation of the interactions being displayed on the side box (8).
  • Lastly, the fourth panel illustrates node importance as determined by the graph neural network (9). Protein residue importance is represented on a blue scale (10), while ligand atom importance is shown on a red scale (11).

Running predictions for mutation effects

User Input Mutations
DDMuffin Mutation Submission Page

The submission page for user input mutations can be accessed via the menu item Run (1) on the top bar menu.

  • The Single Mutation (2) option allows the user to submit one single mutation for prediction.
    • Users are required to provide a wild-type protein-ligand complex structure (3) by uploading a file in PDB format or providing a PDB accession code.
    • Details on the single mutation are should be provided (4) A mutation code consists of a wild-type code, residue position and mutant code (using the one letter amino acid code). The residue position must be consistent with the PDB file. The chain for the mutation must also be provided.
    • The ligand 3-letter code in the PDB should be provided (5).
    • If an email address is provided (6), a link to the results page will be sent to the useras soon as DDMuffin finishes processing the submission.
    • An example of the results page is also provided (7).
  • The Mutation List (8) option allows the user to submit a list of single mutations to be analysed by DDMuffin in batch.
    • Similarly to the Single Mutation option, a wild-type complex structure is required (9) and can be input by uploading a file in PDB format or providing a PDB accession code.
    • The list of single-point mutations must be input (10) as text or by uploading a file containing the mutation codes (chain identifier, wild-type residue code, position and mutant residue code). The file format requires one mutation mutation per line. A sample file is also provided.
    • The ligand 3-letter code in the PDB should be provided (11).
    • If an email address is provided (12), a link to the results page will be sent to the useras soon as DDMuffin finishes processing the submission.
    • An example of the results page is also provided (13).
Pocket Analysis
DDMuffin Pocket Submission Page

The submission page for automated pocket analysis (residues within 5Å distance to the ligand) can be accessed via the menu item Run (1) on the top bar menu.

  • Users are required to provide a wild-type complex structure (2) by uploading a file in PDB format or providing a PDB accession code.
  • On Mutation Details section (3), one is required to provided which type of analysis to run. Here two options are available: Alanine Scanning (mutate all residues within the binding pocket to Alanine) or Saturation Mutagenesis (mutate all residues within the binding pocket to all 19 standard amino acids).
  • If an email address is provided (4), a link to the results page will be sent to the user as soon as DDMuffin finishes processing the submission.
  • Examples of the results pages are also provided (5).

Results for mutation effects

Single Mutation
DDMuffin Single Mutation Prediction page 1 DDMuffin Single Mutation Prediction page 2 DDMuffin Single Mutation Prediction page 3 DDMuffin Single Mutation Prediction page 4

The results page for a single mutation will be shown once computations are completed in 3 different panels.

  • On the first panel, the ΔΔG is displayed (1) alongside with the details on the mutation being analysed (2) and the wild-type affinity prediction (3).
  • The second panel will show and interactive 3D viewer (4), which allows user to analyse the interatomic interactions between the wild-type residue and nearby residues. Interactions can be hidden or displayed using the controls provided (5). The viewer can also be manipulated using buttons at the bottom of the panel (6).
  • On the third panel, DDMuffin allows users to analyse the protein-ligand interactions of wild-type and mutant (7). Here, a set of switches are also provided (8) to allow customisation of the interactions being displayed on the side box (9).
  • Lastly, the fourth panel illustrates node importance as determined by the graph neural network (10). Protein residue importance is represented on a blue scale (11), while ligand atom importance is shown on a red scale (12).
Mutation List
DDMuffin Mutation List Prediction page

The results page for a list of single mutations will be shown once computations are completed in a single table.

  • The predicted ΔΔG for every single-point mutation will be displayed on a table (1). A set of controls for sorting based on columns and an input for text search are also provided.
  • All the analysis discussed for the Single Mutation option can be analysed for each single mutation on the table through the Details button (2) of each row.
  • Results can be downloaded as a .csv file by clicking on the Download button on the top right corner (3).
Pocket Analysis Results
DDMuffin Pocket Result page
Alanine Scanning
DDMuffin Alanine Scanning Results page 1 DDMuffin Alanine Scanning Results page 2 DDMuffin Alanine Scanning Results page 3

The results page for Alanine Scanning will be displayed in three different panels.

  • The predicted ΔΔG for every single-point mutation to Alanine will be displayed on a table (1) on the first panel. A set of controls for sorting based on columns and an input for text search are also provided.
  • Results can be downloaded as a .csv file by clicking on the Download button on the top right corner (2).
  • All results are also summarised on a bar plot on the second panel (3). Positive and negative ΔΔG predictions are shown in blue and red, respectively.
  • The third panel will show and interactive 3D viewer (4) in which the pocket residues are colored according to the ΔΔG predictions. The viewer can be customised using the controls on the top of the panel (5). A set of buttons are displayed at the bottom of the panel (6) to adjust visualisation.
Saturation Mutagenesis
DDMuffin Saturation Mutagenesis Results page 1 DDMuffin Saturation Mutagenesis Results page 2 DDMuffin Saturation Mutagenesis Results page 3

The results page for Saturation Mutagenesis will also be displayed in three different panels.

  • The predicted ΔΔG for every single-point mutation to all possible residues will be displayed on a table (1) on the first panel. A set of controls for sorting based on columns and an input for text search are also provided.
  • Results can be downloaded as a .csv file by clicking on the Download button on the top right corner (2).
  • All results are also summarised on a heatmap plot on the second panel (3) which can be downloaded as an image file (4).
  • The third panel will show and interactive 3D viewer (5) in which the pocket residues are colored according to the average ΔΔG predictions values. The viewer can be customised using the controls on the top of the panel (6). A set of buttons are displayed at the bottom of the panel (7) to adjust visualisation.

Contact us

In case you experience any trouble using DDMuffin or if you have any suggestions or comments, please do not hesitate in contacting us via our Group website.

If your are contacting regarding a job submission, please include details such as input information and the job identifier