GSoC:PTM

BioJava Packages for Identification, Classification, and Visualization of Posttranslational Modification of Proteins

The original proposal is here

Background

Posttranslational modifications (PTM) are modifications to proteins after protein biosynthesis, which play a key role in many cellular processes such as cellular differentiation, protein degradation, signaling and regulatory processes, regulation of gene expression, and protein-protein interactions. PTMs are present in the 3D structures in the Protein Data Bank. It is of vast interst to query and classify proteins by their PTMs. PTMs can be classified in multiple ways. From an implementation perspective we need to distinguish 3 major cases:

  • Case 1: Attachment of a chemical group (i.e. glycan)
  • Case 2: Chemical modification of an amino acid (i.e. hydroxyproline)
  • Case 3: Cross-linking (i.e. disulfide bonds, iso-peptide bonds)

Major Project Contributions

The goal of this project is to develop BioJava packages that identify, classify and visualize PTMs. Major deliverables of this project include:

  • A BioJava package to identify PTMs in a 3D protein structure (.pdb or .cif file).
  • A BioJava package to generate sequence diagrams with an option to add PTM annotations.
  • A BioJava package to generate 2D tree images of carbohydrate (glycan) structures.

Immediate applications of this project would be on Protein Data Bank website:

  • Making PTMs searchable on PDB.
  • Updating the sequence diagram on PDB with an option to display PTM annotations.
  • Listing PTMs in PDB ProteinWorkshop.

This project will be beneficial to the research community by facilitating structural analysis on PTMs and hence reinforcing our understanding on the mechanisms of various PTMs.

Tasks and Implementation

Task1. Making a list of PTM types

Resources

Procedure to get a list of PTMs

  • Retrieve the Chemical Component Dictrionary
  • Remove all obsolete or ambiguous chemical components by only keeping entries with
    • _chem_comp.pdbx_release_status REL
    • _chem_comp.pdbx_ambiguous_flag N
  • Create two subsets of the CCD
    • A. All chemical components that are
      • not the 20 standard amino acids and
      • do not have a _chem_comp.mon_nstd_parent_comp_id field corresponding to one of the 20 standard amino acids.
    • B. All chemical components that have a _chem_comp.mon_nstd_parent_comp_id field corresponding to one of the 20 standard amino acids.
  • For case 1 (attachments), match the RESID NameXref PDBHET ID with chemical component set A to get a list of attachment groups.
  • For case 2 (modified amino acids), match the RESID NameXref PDBHET ID with set B to get a list of modified amino acids that are PTMs.
  • A initial list of PTMs were then manually selected.

Task2. Defining data representation of PTMs

  • An XML file is used to store the information of PTMs.
  • A Java class ProteinModification to store different types of PTMs.
  • Three interfaces representing three cases.
    • ModifiedResidue
    • Attachment
    • CrossLink

Task3. Reading and parsing 3D protein structure files (.pdb or .cif)

  • Utilizing org.biojava.bio.structure (http://www.biojava.org/wiki/BioJava:CookBook:PDB:read).
  • The code will be based on BioJava 3.

Task4. Identifying PTMs in 3D protein structures

  • For case 1, finding the attachment points for PTMs.
    • Reading modifications in HETATM.
    • Scanning possible atoms on possible amino acids to locating attachment point of each PTM by checking if the distance between an amino acid atom and the PTM is less than a threshold.
      • The threshold will be the sum of the covalent bond length (i.e. the sum of covalent radiuses of both atoms) plus a tolerance of error, say 0.4 Angstrom (this need to be decided later after analyzing the data).
      • For different types of PTMs, only limited types of amino acids and atoms can be attached to. Thus, only those atoms need to be scanned.
      • If multiple atoms in multiple amino acids are within the distance threshold to the PTM, choose the one with the shortest distance.
  • For case 2, finding the modified amino acids.
    • Finding the corresponding 3-letter code of each PTM type in Chemical Component Dictrionary.
      • If _chem_comp.mon_nstd_parent_comp_id field corresponds one of the 20 standard amino acids, then it is a modified amino acid.
      • 3-letter code is contained in the _chem_comp.id field.
    • Identifying PTMs by parsing through the residues in the protein chain using the 3-letter code.
  • For case 3, finding cross-linked amino acids.
  • For both case 1 and 3, filtering out the close contacts that have nothing to do with PTMs.
    • The filtering strategy will be developed after analyzing the data.

Task5. Representing PTMs in text

  • For case 1: Attached PTM
    • Two atoms that link between the amino acid and the attachment
      • 3-letter code
      • Chain id
      • Residue number
      • Atom name
      • Distance between the two atoms
      • Example:
        • ASN A 173 ND2
        • NAG A 651 C1
        • 1.45 A
    • Two atoms that link between additional chemical components
      • 3-letter code
      • Chain id
      • Residue number
      • Atom name
      • Distance between the two atoms
      • Example:
        • NAG A 651 O4
        • NAG A 652 C1
        • 1.43 A
  • For case 2: modified amino acid
    • 3-letter code
    • Chain id
    • Residue number
    • Example:
      • HYP A 123
  • For case 3: cross-link
    • 3-letter code
    • Chain id
    • Residue number
    • Insertion code
    • Atom name
    • Distance
    • Example
      • LYS A 36 NZ
      • ASN A 168 CG
      • 1.5 A

Task6. Annotating PTMs on sequence diagram view

  • Refactoring the Java classes that are being used in PDB to diaplay the sequence diagram into stand-alone Java classes and make it available in BioJava.
    • In the sequence diagram of an entry in PDB (e.g. 3M6S), a user can select information/annotations (e.g. Pfam domain, InterPro domain, etc) of one’s interest.
    • When refactoring, a key issue is to keep the flexibility to add new annotations, such as PTM annotation.
  • Extending the diagram with an option to display PTMs in the structure.
    • For attached PTMs (case 1) and modified amino acids (case 2), place triangles and abbreviations over the residue.
      • Whether to display the PTM annotation is an option to the users, just like other annotations.
      • Different colors of the triangles and texts could be used to distinguish case 1 and case 2 PTMs.
  • For cross-link, linking the two residues with a dotted line.
    • Currently in PDB, disulfide bonds is displayed without an option.
    • A better way will be leaving the cross-links as an optional annotation to users.
      • Different types of cross-links (say disulfide bond and iso-peptide bond) will have separate options.
      • Different line color or style (dotted, dashed, etc) could be used to represent different cross-link types.
  • Adding another option on the sequence diagram to display PTMs annotations from UniProt (or other sources).
    • UniProt/Swiss-Prot contains tens of thousands of PTM annotations, which can be displayed in the sequence diagram.
    • The UniPort XML file (e.g., http://www.uniprot.org/uniprot/P56524.xml) can be utilized.
    • Some code I have written in Musite could be refactored and used here.

Task7. Generating 2D tree images of glycan structures

Timeline

  • 05/24-05/25: Task 1-Making a list of PTM types.
  • 05/26-06/02: Task 2-Defining data representation of PTMs.
  • 06/03-06/09: Task 3-Reading and parsing 3D protein structure files (.pdb or .cif).
  • 06/10-06/23: Task 4-Identifying PTMs in 3D protein structures.
  • 06/24-06/25: Task 5-Representing PTMs in text.
  • 06/26-07/30: Task 6.1-Refactor sequence diagram code from PDB.
  • 07/31-08/06: Task 6.2-Annotating PTMs on sequence diagram view.
  • 08/07-08/16: Task 7-Generating 2D tree images of glycan structures.

Weekly Report

  • 05/31/2010
    • A initial list of case 1 and case 2 PTMs has be identified.
    • An XML format to store the information of PTMs has been defined.
    • Java classes representing the PTMs have been committed to the code repository.
  • 06/07/2010
    • Case 3 PTMs were added into the XML file.
    • XML parser.
    • Identification of case 1 PTMs.
    • Unit tests.
  • 06/14/2010
    • Identification of PTMs of all three cases.
    • More modifications were added into the XML file.
    • Representing PTMs in text in test code.
  • 06/22/2010
    • More modifications were added into the XML file, such as isopeptide bonds.
    • Improving the current code according to
    • PDB sequence diagram code was refactored from PDB.org by Andreas.
  • 06/28/2010
    • Finished going over all RESID entries.
      • 212 entries were added in the XML file.
      • Special treatment should be applied to all ACE entries.
    • Identify additional attached ligands that are not directly attached to amino acids.
    • List unidentifiable atom linkages.
  • 07/08/2010
    • In progress of scanning whole PDB.
    • Hacking the PDB sequence diagram code.
  • 07/16/2010
    • Attending ISMB from July 09-13
    • Finished scanning whole PDB. Need to analyze the result.
    • Started to refactor the sequence diagram code.
  • 07/26/2010
    • Keywords added in the XML file.
    • Analyzed results of PDB scanning: summary.
    • Still working on sequence diagram code.
  • 08/02/2010
    • Refactor ModifiedCompound and ModifiedCompoundImpl.
    • Serialization.
  • 08/09/2010
    • Refactor package structures
    • Integrate crosslinks in sequence diagram.
  • 08/16/2010
    • Better representation of PTMs in sequence diagram.
    • Support metal coordination modifications.

Skype call notes

  • 07/29/2010
    • Participants: Jianjiong, Peter.
    • Top priority: add PTM annotations in sequence diagram.
    • Serialization: using Strings to represents identified PTMs.
    • Metal Coordination
      • classify according to number of residues involved
      • some metal ions links to multiple residues forming cross links, e.g. AA0136
    • Unidentified modified residues: contact John Garavelli
    • Do we need to scan different models in a structure? No, use the first model.
    • Improve Java doc
    • Add a chapter in cookbook: http://biojava.org/wiki/BioJava:CookBook3.0
  • 06/15/2010
    • Participants: Jianjiong, Peter, Andreas.
    • Find more test cases for cross-links (at least one for each type of cross-link)
    • Scan whole PDB to find more modifications
      • Classify non-natural modifications into the same classes as the natural modifications (attachments, modified residues, cross-links)
      • Alternative confirmation (altloc, see PDB Content Guide ) could be a reason for close contacts.
      • Send to Andreas if errors occur when reading structures.
    • How to deal with multiple modifications (in Green fluorescent proteins), e.g. CRO in 3MIQ?
    • About code
      • Break long functions (e.g. in DefaultProteinModificationParser) into short ones.
      • Collections.singletonList for 1-element list.
      • Return empty collection instead of null.
      • Use FindBugs plugin to detect potential bugs
      • Remove any Eclipse/Java warnings if present
      • Follow Sun JavaDoc conventions 1
      • Any Checkstyle template for BioJava?
    • Cookbook page after API is stable.
    • Andreas is refactoring sequence diagram code from PDB web.
    • In sequence diagram, dashed line is only for cross-link2. How to represent cross-links that link more than 2 residues?

Comments

Please add comments here…

  • Peter Rose (04/30): It’s an ever increasing list of PTMs. So instead of hardcoding PTMs, it would be better to load them from a file, i.e. xml.