Protein Structural Bioinformatics Course

Syllabus for
Protein 3D Structure Visualization & Structural Bioinformatics

Part of the Applied Molecular Biotechnology MS Program, Dept. Microbiology,
University of Massachusetts, Amherst.
2019: Mondays and Thursdays, January 7, 10, 14, 17 in Fine Arts Center 444.
1:00 to 4:00 PM each day.

Bringing A Laptop Computer Is Encouraged
but not required (lab mac & windows computers are available).

Taught by Eric Martz* and Jeffrey Kane.
*principal author of FirstGlance in Jmol and MolviZ.Org,
team member of Proteopedia.Org, and coauthor of the ConSurf Server.
*Professor Emeritus, University of Massachusetts, Amherst -- Morrill IVN 422B --

This syllabus is on-line:
Workshops.MolviZ.Org

Goals: This course will prepare students to understand and incorporate 3D macromolecular structure into their research and teaching. The principles of protein structure will be reviewed, including noncovalent bonds. Structural bioinformatics and genomics will be introduced. Students will learn what percentage of proteins have known 3D structures, and the importance of crystallographic models compared to homology models, or theoretical models.

Using laptop computers, students will learn how to find 3D protein molecular models for proteins of interest, or how to construct homology models,

and how to use the FirstGlance in Jmol 3D visualization software to investigate key structural features.

Protein structure will be related to function, evolutionary conservation and multiple-sequence alignments, and drug design. Specific oligomers will be constructed and visualized. Students will learn how to prepare customized publication-quality molecular images and animations for Powerpoint slides. Each student will prepare a report, using Powerpoint slides to capture the concepts and skills they have learned. All the software is web browser-based, easy to use, works on Windows or Mac OS X, requires no installation, is free and open-source, and is expected to be available for years to come.

Get Started

Each student please get started:

If you brought your own laptop, you are welcome use it. (iPads will be too slow.) You may use Mac OS X or Windows, whichever you prefer. The software works the same in both.
Use the Firefox or Chrome browser. If you do not have either, take a few minutes to install Firefox. Chrome rotates molecules a little more smoothly than Firefox, but Firefox loads and displays larger molecule much faster than Chrome. If your molecule takes more than one minute to display in FirstGlance, ask the instructor for guidance: Using Java will be much faster.

Go to our syllabus: Workshops.MolviZ.Org.
Now you can see this document in your browser. Go to Atlas.MolviZ.Org.
In the Atlas, choose any molecule deemed Straightforward and click on the link to FirstGlance. After a minute or so to load, you should see a rotating molecule. Have a look around at the information, views and tools in FirstGlance.
If you have any difficulty or the molecule does not appear, or does not rotate, ask for help!

Workshop Startup and Overview

Workshop Startup & Overview (Powerpoint Slides)

Nikhil Malvankar (Physics) took this workshop in 2012. Published his homology model in 2015.

I. Protein Data Bank & PDB Codes
Crystallographic Resolution
Proteopedia.Org

The Protein Data Bank (PDB) -- World Wide: USA:RCSB -- Japan:PDBj -- Europe:PDBe. All 3 have the same data.
- Primary vs. Derived 3D Macromolecular Structure Databases
- What is a PDB file?
PDB identification code examples:
- 1pgb a small protein domain, one chain.
- 2hhd four protein chains with ligands.
- 1d66 protein and DNA.
- 104d DNA/RNA hybrid.
- 9ins protein hormone.
Proteopedia.Org
1. Main page: green links connect text to molecular scenes.
2. Molecules explained by users. Examples:
  - Authored by UMass grad students: Molecular Playground/DHFR
  - Other UMass student pages: CBI Molecules
  - Well developed by 9 authors: Avian Influenza Neuraminidase, Tamiflu and Relenza.
3. Explanations of structural biology terms and concepts, e.g. asymmetric unit, Protein Data Bank, hydrogen bonds, temperature value, etc. all at About Macromolecular Structure.
4. Pages in Spanish, Chinese, Russian, Arabic, Japanese, Turkish, etc.
An electron density map at 2.5 Å resolution.
X-Ray Crystallography and Resolution
- 85% of models in the PDB come from X-ray crystallography experiments.
- X-ray crystallography produces an electron density map (EDM).
- The average uncertainty in an EDM is measured by its Resolution in Ångstroms:
  - 1.2 Å Excellent -- backbone and most sidechains very clear. Some hydrogens resolved.
  - 2.0 - 2.5 Å Good -- backbone and many sidechains clear. Typical X-ray crystallography.
  - 3.5 Å OK -- backbone and bulky sidechains mostly clear. Typical cryo-electron microscopy.
  - 5.0 Å Poor -- backbone mostly clear; sidechains not clear.
  - See the MOVIE.
- The success rate in crystallography is overall about 5%.
- See also Quality Assessment for Molecular Models, R value and especially Free R.
Electron Cryo-Microscopy is increasingly important
- Crystallization, the main bottleneck in X-ray determination, is unnecessary.
- Structure can be determined for some assemblies without cloning/purification/expression. In some cases, the protein is first identified from the structure. Example.
- 3 min video explanation of method (silent).
- Good-resolution X-ray crystallographic models can be docked into poor-resolution cryo-EM maps of large assemblies. Example (alt link).
- Near atomic resolution can sometimes be achieved. 5FTJ 2.3 Å. Median 3.8 Å.
- 2017 Nobel Prize in Chemistry.
- Still not quite 2% of all entries in the Protein Data Bank.

II. Obtaining models for molecules of interest.
Begin Powerpoint Slides.

Finding molecular models of interest:
Each student: please find a 3D model of a molecule related to your research or interests.
You will use the model that you select for the rest of the class, and for your Powerpoint report.
- PDB code for an X-ray crystal structure with good resolution <= 3.0 Ångstroms.
- Has protein.
- Has ligand(s).
- Has R_free (free R).
  
  If An Empirical Model Is Not Available
- Homology model. See Limitations of Homology Modeling.
- Use a homology model for your report only if there are at least 120 amino acids in one chain. Smaller homology models usually represent less than a single domain and are not useful.
- Don't use a homology model if an empirical (X-ray or NMR) model exists for your protein. The empirical model will be more accurate. If unsure what to use, please ask the instructor.
How To Find Models:
- Go to UniProt.Org.
- Find your molecule. Ask for help if needed.
- Click on the blue Sequences button at the left.
- Make a note of the length of the full-length amino acid sequence.
- Click on the blue Structure button in the left column. (Examples: yeast gal4; human pla2g6; zebrafish acetylcholinesterase; human p53.)
- Follow instructions for the Simple search, and if necessary the Advanced search, at
  Is There An Empirical Model?. ("Empirical" means determined by X-ray or NMR.)
- For any model that you find, pay attention to how much of the full-length sequence it covers.
- If there is no empirical model, is there a homology model?
  See How To Obtain Homology Models.
- For any model that you find, pay attention to how much of the full-length sequence it covers.
- To use a homology model for your report, submit the sequence to generate a new model. This will give you the most up to date results, and the best sequence alignment listing. (You cannot get a sequence alignment easily from ModBase.)
  1. Copy your amino acid sequence (e.g. from UniProt).
  2. Go to Swiss-Model.
  3. Click Create Account or Log in.
    It is worth bothering to create an account because it makes your jobs easy to find later. Without an account you might have to re-run the job and it can take over an hour.
  4. Click the blue Start Modeling button near the top center, or use the pull-down Modelling menu and click on Automated Mode.
  5. Paste your sequence in the large box.
  6. Click the large blue button Build Model.
- After Swiss-Model makes a model for you, download the model PDB file (see snapshot at right).
- Upload your homology model PDB file to FirstGlance:
- Use a homology model for your report only if there are at least 120 amino acids in one chain. Smaller homology models usually represent less than a single domain and are not useful.
Browsing: If you can't find a model for your protein, or you don't have a molecule in mind, look at one of these sites and pick one.
- Atlas of MacroMolecules (Atlas.MolviZ.Org)
- Molecule of the Month
- Structure Index or Topic Pages at Proteopedia.
Begin your Powerpoint Slides (Later, you will email them to Prof. Martz)
- Complete questions 1-3 now: Questions for your slides.

III. Review of Protein Chemistry and Structure.
Introduction to Structural Bioinformatics.

Standard amino acids.
Click to see details.

Central Dogma: DNA mRNA Protein. DNA structure in Jmol / Estructura del ADN
20 Amino acids

Codon = 3 nucleotides; 4 nucleotides³ = 64 codons.

3-letter and 1-letter codes.
Amino acids in Jmol
Chemical properties (Properties; Venn Diagram)

Polar, uncharged (cf. water)
Charged
Hydrophobic (aliphatic vs. aromatic)

Polypeptide chain geometry and steric restrictions

Ramachandran Principle, Phi (φ) & Psi (ψ) angles in amino acids
also in a narrated YouTube Video.
Gly (turns), Pro (turns, helix breaking: 245-277 in 1f88).

Covalent and non-covalent chemical bonds

Typical hydrogen bond within a protein: hydrogen donor atom is covalently bonded to hydrogen; acceptor atom is not. In proteins, donor-acceptor distance can be 2.5 to 2.5 Å.

Covalent bonds: lengths and angles nearly constant
Non-covalent bonds: variable lengths and angles

Salt bridges (up to 4.0 Å in proteins)
Hydrogen bonds (2.5-3.5 Å donor-to-acceptor in proteins)
Cation-pi interactions (up to 6.0 Å in proteins)
van der Waals interactions (up to 4.0 Å in proteins)

Primary, Secondary, Tertiary, Quaternary Protein Structure
Los cuatro niveles estructurales de las proteínas
Secondary Structure

Backbone traces and smoothed ribbons.
Chains vs. Domains. Example: 1IGY.
Alpha helices in Jmol

Right handed (with L amino acids). C=O receives hydrogen bond from N-H of 4th residue ahead.

Beta strands and sheets in Jmol
Turns (Glycine, Proline), Breaks in helices (Proline)

Tertiary Structure: Folding & Hydrophobic Collapse

Protein folds cannot be reliably predicted from sequence alone (using ab initio theory).

Quaternary Structure

Biological Units
Four Levels of Protein Structure, an interactive, discovery-based tutorial with built-in reporting and assessment, which is part of Molecular Workbench from the Concord Consortium.

Intrinsically Disordered Protein

Fold does not always equal function: About 10% of proteins are thought to be fully disordered to support their functions, and 40% of eukaryotic proteins have at least one long disordered region. This is termed intrinsic disorder. Examples: 1jsu, 2rrl.

Introduction to Structural Bioinformatics
- Why do we care about 3D macromolecular structure?
- What are 3D structure data?
- Where do 3D structure data come from?
- How much 3D structure knowledge do we have?
- Primary and Derived 3D Structure Databases

IV. FirstGlance in Jmol for exploring any macromolecule.

FirstGlance in Jmol (Part I).

To start FirstGlance, google "firstglance" (no space), or go to FirstGlance.Jmol.Org. Enter the PDB code, or upload your homology model.

Unusually large models may take a long time to display and be sluggish to manipulate in FirstGlance. For such cases, using Java will enable much better performance in FirstGlance. Java is not needed for most models. Ask the instructor for advice. Here are instructions for Installing and Enabling Java.
Explore 1izh in FirstGlance.
1. Introduction
2. Molecule tab
  1. Year, Method.
  2. Resolution.
  3. Free R.
  4. Chain details.
  5. Sequences: Crystallized vs. Full Length. Alignment at UniProt (1d66).
  6. Abstract.
  7. Citations.
  8. Text contents of the PDB file.
3. Views tab
  1. Top 3 rows of views:
    Secondary Structure / Cartoon / N->C Rainbow
    Composition / Hydrophobic/Polar / Charge..
    Local Uncertainty / Vines / Thin Backbone
  2. Buttons.
    Ligands+ / Water / Slab
  3. 1pgb: Hydrophobic core: Hydrophobic/Polar, then Slab.
  4. 1pgb: Amphipathic helices and strands. (In FirstGlance, use Isolate.. on each end of a helix or strand.)
  5. Compare with the Hydrophobic/Polar View of 1bl8 or 7ahl.
4. Resources tab
  1. See lipid bilayer boundaries (1bl8 or 7ahl).
5. 1pgb: Tools tab with Views.
  1. Salt bridges.
  2. Cation-pi interactions.
  3. Distances.
  4. Salt bridges in Charge View. (Red sidechain (-) touching blue sidechain (+)).
  5. Charges with Slab on.
  6. Sidechain distributions in Vines View (rings buried; charges on surface).
  7. Find (review Chart of AA): PHE, (VAL,LEU,ILE), ASN, THR
Explore 9ins in FirstGlance.
1. Tools tab
  1. Disulfides/S/Se
Explore 3onz in FirstGlance. (Letter O not numeral zero!)
1. Molecule Information Tab
  1. Two chains, not sequence identical.
  2. Missing residues.
  3. Ligands+ and non-standard residues
2. Views tab
  1. Ligands button; smaller ligands.
  2. Hide (chain, toluene, isolated His).
3. Tools tab
  1. Contacts
  2. Non-covalent interactions for HEM in chain A (blue chain).
4. Resources tab
  1. Biological unit.
Solution Nuclear Magnetic Resonance (NMR)
- Gives an ensemble of multiple models consistent with the data.
  Examples: 1abt, 1cfc, 1jsa; 2fft (intrinsic disorder).
- Differences between models can reflect flexible thermal motion in solution, or simply uncertainty due to a lack of enough data. Nothing in the PDB file tells you which is the case. You need to contact the authors.
- There is nothing in the PDB file that measures reliability. (Unlike X-ray data, where Resolution, R, and R-free measure reliability.)
Continue preparing slides to answer the Powerpoint Questions.

V. Introduction to Multiple Sequence Alignment (MSA) and Conservation
ConSurf Server
Structure of Atomic Coordinate ("PDB") Files

		Enolase (4enl; a glycolytic enzyme) evolutionary conservation from ConSurf. Catalytic cleft is at the left. More..

Identify Functional Sites In Your Molecule Using The ConSurf Server:

Go to the ConSurf Server.
Select Amino Acids.
Check YES there is a known protein structure.
Enter the PDB code, or upload your homology model.
Select a chain for analysis. (If your PDB file has only one chain, it will be selected automatically.) ConSurf can analyze only one chain at a time.
Check NO, you will not be uploading a multiple sequence alignment. (ConSurf will create one for you.)
Leave all search parameters at their pre-set defaults.
Select sequence homologs for ConSurf analysis:
Check manually for a research-grade result.
Check automatically for a quick but probably lower quality result.
Leave the alignment method, calculation method, and evolutionary substitution model at their default settings.
Enter your email address at the bottom of the form (this is important so you don't lose your results). Optionally enter a job title.
Click the button Submit. It may take an hour or more to complete this calculation.
If you are selecting sequences manually, you will see SELECT SEQUENCES.
For instructions, see Limiting ConSurf Analysis to Proteins of a Single Function.
Example: Sequences ready for selection (2vaa, Swiss-Prot).
When it is finished, under Final Results, click the link View ConSurf Results with FirstGlance in Jmol.
Example: 4enl result in ConSurf.

Background: see Introduction to Evolutionary Conservation.

Effect of mutation on protein function	Genetic consequence	Example
Function LOST**	CONSERVED: mutation LOST from gene pool	R133C*
None	NOT conserved: mutation remains in gene pool	E143?*
* in methyl CpG binding protein 2 (MeCP2), 3c2i: `ASASPKQRRS IIRDRGPMYD DPTLPEGWTR KLKQRKSGRS AGKYDVYLIN PQGKAFRSKV ELIMYFEKVG DTSLDPNDFD FTVTGRGSPS RHHHHHH ^ ^` ** R133C causes Rett syndrome, a severe neurological disorder. Gray: disordered in crystal, absent in model 3c2i.

Enzyme example: ConSurf-colored sequence -- 4enl ConSurf Result -- enolase in Wikipedia.
Multiple sequence alignments reveal conservation: MSA for 4ENL in black and white.
Detail of MSA. ConSurf does a much more sophisticated job of calculating evolutionary conservation scores than this simple example!
On the less conserved side of the molecule, touch the isolated highly conserved residues to display the amino acid and sequence number. Gly236 and Pro290 are highly conserved. Why?.

ConSurf Mechanism. (Details of Mechanism).
Note the Caveats in Proteopedia's Evolutionary Conservation.

Atomic Coordinate Files
- Formats
  - Crystallographers: "PDB format" (Human-readable; based on 1970's system that used 1928-design paper punch cards)
  - Protein Data Bank: mmCIF (macromolecular crystallographic information format)
  - US National Center for Biotechnology Information: ASN1
Continue preparing slides to answer the Powerpoint Questions.

VI. FirstGlance in Jmol -- Part II
Solution NMR
Isoelectric Point
Intrinsically Unstructured Proteins

FirstGlance in Jmol (Part II)

Solution Nuclear Magnetic Resonance (NMR)
- Gives an ensemble of multiple models consistent with the data.
  Examples: 1abt, 1cfc, 1jsa.
- Differences between models can reflect flexible thermal motion in solution, or simply uncertainty due to a lack of enough data. Nothing in the PDB file tells you which is the case. You need to contact the authors.
- There is nothing in the PDB file that measures reliability. (Unlike X-ray data, where Resolution, R, and R-free measure reliability.)
Charge:
- 1d66 and nucleotide binding.
- Challenge: how can protein charge be changed in seconds, without changing the pH? .
- Calculate the isoelectric point (pI) and charge at pH 7 for one chain of your protein:
  1. Follows the instructions under Charge.. in the Views tab.
Intrinsicially Unstructured / Natively Disordered Proteins

VII. Publication Quality Images and Animations with Polyview-3D
Finishing Powerpoint Questions

Animation from Polyview-3D.
Click on the above image for
a larger view and explanation.

Animate customized molecular scenes easily with Polyview-3D.

Animations are most easily saved from FirstGlance, but you are limited to the "canned" color schemes and renderings that it offers.

If you wish to customize a molecular scene with arbitrary colors and renderings, there are two relatively easy options:
1. Proteopedia.Org has molecular scene authoring tools: you use checkboxes and forms to customize the scene, and it is immediately online to share. It is easy to make a presentation-ready animation from your customized Proteopedia scene.
2. Polyview-3D also allows customization of colors and renderings from forms and checkboxes. Its output is a high-quality animation (rendered with PyMOL) or publication-quality image.
Once you have an animation from Polyview-3D:
- Windows Powerpoint: Simply drag the animation directly from the Polyview-3D web page and drop it into a powerpoint slide.
- Mac OS X Powerpoint: (This works in Powerpoint:Mac 2008 and Powerpoint for Mac Version 15.41.)
You are now prepared to finish your Powerpoint Questions. Please email the completed PPT file to
emartz AT microbio DOT umass DOT edu.

Additional Resources.
Probably we will not have time in class to spend on these resources. Links are provided here in case you are interested to look at these later.

Simplified SV40 Virus Capsid.

Lac repressor bending the DNA operon. If this image is not moving, reload the page.

Animations & Morphing
Mutating your model:
1. Changing residue sidechains and rotamer minimization with DeepView.
2. DeepView beginners should start with the superb Molecular Modeling for Beginners by Gale Rhodes, Univ. Southern Maine.
Structural Alignment of two or more chains or molecules, and how to view the alignment.
- DeepView can align macromolecules (one or more than one chains), selecting any subset of atoms for the alignment (other atoms following), and retaining hetero atoms. The results can be saved as a PDB file, but will need manual editing to separate models with MODEL [N] and ENDMDL records so that Jmol can distinguish the models. Gale Rhodes provides a DeepView tutorial: click on the section Comparing Proteins.
- See also Structural Alignment Tools.
For Teachers and Future Teachers
High School Teacher's Resources.
MolviZ.Org
1. DNA (with quiz), Hemoglobin (with quiz)
2. Atomic clashes: Ramachandran Principle (animated, interactive, with quiz)
3. Lipid Bilayers and Gramicidin Channel
4. Collagen
5. Water & Ice & hydrogen bonding
6. Toobers in Science Education
About Macromolecular Structure
Building a web page that shows your favorite molecules for research or teaching.
- Proteopedia is by far the easiest. Your text and molecular scenes can be protected and are online and searchable immediately!
- Making hyperlinks that show your molecules in FirstGlance in Jmol.
- Other authoring tools for web pages with Jmol

Keep in touch!

Don't hesitate to email me with any questions that you have: emartz@microbio.umass.edu, or we can make an appointment at UMass Amherst (or by Skype).