Model

Below you can find our model's assumptions, data, parameters, and results along with the tools used and our reasonings.

Assumptions:

  • We ran all computational tests with non-biotinylated peptide and DNA aptamers despite ordering biotinylated aptamers for our wet lab tests - we did this under the assumption that the biotin doesn’t affect the binding affinity much
  • We were unable to find a pdb file for the full ferritin multimer. We tested binding affinities between our aptamers and the light and heavy chains of ferritin. We assumed that if our aptamers are able to have a decent binding affinity with these single chains, then the binding affinity will be even better with a macromolecule made up of many of these chains. We are ignoring sterics by doing this, however, so that is something to take into consideration when doing wet lab tests.
  • We’re also assuming that the aptamer won’t bind to ferritin multiple times once presented with more than one chain to bind to. This assumption is being made off of the data we found in our literature review and that the thermodynamics of the system will prevent the ligand and protein from binding multiple times.
  • While there are a lot of tests backing the peptide and DNA’s binding affinity with ferritin in the papers and they give a decent amount of information on the testing conditions, we don’t have all the details (pH, temperature, pressure, etc) needed to properly simulate the experiment. All of the computational docking tests were run at STP and a pH of 7. This is also good for our biosensor as we hope for it to not need refrigeration.
    • As an add-on to that, we are assuming we can store the final biosensor at room temperature. The peptide and DNA we ordered was pretty stable but genscript did still give lower than room temperature storage conditions for both. We are assuming that the aptamer structures being as they are, especially with DNA being generally stable, they can last at room temperature for testing at least if not general storage.
  • For potentially using SimRNA and other sites that give pdb files of RNA secondary structures, we are assuming that RNA will have a similar binding affinity with ferritin as DNA does.
    • In other words, the change from T to U won’t make much difference in binding affinities (still need to verify this, and our first option is still using the oligo single strand DNA instead of RNA)

Tools that were used (protocols, results, details on why they were used):

GROMACS

    We initially tried using GROMACS for ligand-protein binding. We weren't able to, however, as we couldn’t find any full pdb files for ferritin (some residues missing from all files found in protein data bank). We ended up using GROMACS to clean the pdb files and energy minimize the ligands. We mostly followed the GROMACS tutorials to do the protein-ligand complex. Khalid, one of our captains, also guided us on how to use GROMACS on the NYU HPC specifically. This included showing us how to upload files to the HPC, the syntax differences, and how to run batch commands.

Autodock Vina

    We ended up using Autodock Vina for testing the ligand-ferritin complex. For this software, we just had to input pdb files for both the ligand (peptide aptamer) and macromolecule (ferritin - light or heavy chain), and we did not run into any issues because of one or two missing residues on the ferritin files. Below is the protocol we used. It was developed by Emre Erkanli, a PhD student in Dr. Kim’s lab who uses a lot of computational docking for his research.

    Protocol (Developed by Emre Erkanli) - Paraphrased and Shortened:

    • Open the protein and ligand in Pymol one at a time - clean up (delete unnecessary chains for protein) and save as pdb
    • Open Autodock → Read Molecule → open protein/enzyme (ferritin in our case)
      • Edit → delete water molecules
      • Edit Hydrogen → selection options: polar, non-bond order, yes
      • Edit Charges → add Kollman charge
      • Save edited protein as a macromolecule (pdbqt)
    • Open the peptide.pdb as a ligand input and save it as peptide.pdbqt (ligand output)
    • Open the gridbox → adjust spacing to 1.0 → adjust box dimensions until it covers the entire protein (macromolecule) → output grid file as grid.txt
    • Create text file “config.txt” and add which files will be receptor (protein) and ligand (peptide), dimensions (size, center coordinates) of the gridbox from grid.txt, and the energy range and exhaustiveness of the program
    • Run Autodock vina - select file names vina and config.txt for two inputs → run program
    • Command screen should pop up showing progress → will get a table of binding affinities and rmsd’s of the different ligand conformations
    • Can save output file as a pdb and open to view ligand conformations with the macromolecule in Pymol and/or Autodock Vina

    Here is the first config.txt we made (ligand that was only energy minimized in Chemdraw + Heavy Chain Ferritin)

    conf.txt

    Here are the results we got on the command screen

    cmd

    Here is the first ligand conformation with ferritin

    autodock

NYU High Performance Computer

    We used NYU’s HPC mainly for GROMACS. The plan was to just submit computational docking tests to the HPC so we wouldn’t be taking up a lot of time on our personal computers. However, we ended up switching to Autodock due to the issues we were facing with GROMACS.

    We tried adding Autodock onto the HPC to somehow use it there as each Autodock program ran about 6 hours on a personal computer. We ran into issues, however, getting the Autodock software onto the HPC, as we couldn’t figure out a way to download the software into the HPC. We’re now looking into using Open OnDemand (opening their computer desktop on mine) to install Autodock and add all of our files there. This can cut down on a lot of time if we continue with computational docking for the peptide-ferritin complex. It’s not as much of an issue for the DNA-ferritin tests as Haddock (details below) is a web server that runs its program in the cloud instead of on the computer.

PyMol

    PyMol was mainly used to open pdb files of the peptides and ferritin in order to view and clean them (see Autodock protocol above for more info). We also tried using PyMol initially to draw the peptide as the software comes with an amino acid building blocks option. The structure wouldn’t maintain after the first few amino acids, however, as it was building somewhat helically and the sterics weren’t allowing us to add more amino acids. There was no real protocol for this as it was just us messing around to see what can be done. Obviously, it didn’t work, so as a result, Amulya drew the peptide organic structure on chemdraw and energy minimized using Chem3D.

    Additionally, we tried making the DNA aptamer on PyMol using base pair building blocks. Unfortunately, we were only able to make a helical structure, not a proper secondary structure. When we later tried energy minimizing using Haddock (see below), the structure remained helical. We know this because the conformations we got from the computational docking tests still had a spiral shape and were only slightly bent. We might still use this spiral structure from PyMol and input it into other energy minimization softwares to see if any of them work. This is only if we can’t find any other simple options for getting a pdb file of the DNA secondary structure.

Chemdraw

    We used Chemdraw to draw the ligand (peptide aptamer) from scratch after the PyMol amino acid building blocks weren’t working. We also used it to do an initial energy minimization of the peptide, but the structure was still very stiff, indicating that the energy minimization didn’t work very well.

    Here is an image of our peptide after it was drawn and energy minimized using Chemdraw.

    no_em_chemdraw

Chem3D

    We used Chem3D to do another energy minimization on the peptide as Chemdraw’s EM was not very good. This time, the structure came out much more bent in shape and the Autodock results were a bit better when we retested the peptide with heavy chain ferritin (aka the magnitude of the binding affinity we got for ligand-ferritin complex was a bit bigger).

    Here is an image of our peptide after the Chem3D energy minimization in Chem3D.

    em_chem3d

    Here is an image of our peptide after the Chem3D energy minimization in Chemdraw.

    em_chemdraw

Haddock

    Haddock is an online software that we used to energy minimize the single strand DNA aptamers and test the binding affinity of the double strand, 5’-3’ single strand, and 3’-5’ single strand DNA aptamers with ferritin (light and heavy chain). From our literature review, it appeared that the DNA can bind to both heavy and light chains, so we wanted to confirm.

    As we were unable to make a good secondary structure for the DNA using Pymol, we tried energy minimizing the DNA using Haddock (as mentioned earlier) to try getting a correct secondary structure. After running the binding test between the three DNA aptamers listed above and the two ferritin chains, we saw that the conformations of DNA were not very different from the original spiral shape. In other words, the Haddock energy minimization didn’t really work.

    Unfortunately, the images from this test were not saved properly and the files were corrupted. When we went back to take photos again, we saw that the online tool automatically deletes results from your account after a while. We are currently in the progress of getting a pdb file of the actual DNA secondary structures to test with ferritin. Hopefully, when we run the Haddock test with better DNA structures, we will get better conformations and results to present.

All of the DNA Secondary Structure Tools

    We’re currently trying to find if there are any other softwares that can create a pdb or sd file of the correct DNA secondary structure. We’ve unfortunately had no luck so far. Our next steps are to either:

    • Try energy minimizing DNA aptamers (5’-3’ strand and 3’-5’ strand) using other softwares like Chem3d to see if this can maybe provide a better secondary structure.
    • Work on just writing the DNA aptamer organic structure from scratch. This will take a while and require a better understanding of how secondary structure linkages and bonds work.

    Vector Builder

      Vector Builder is an online tool that creates a secondary structure png output for a DNA sequence input. It was used to get a secondary structure of DNA. Using this, we had a dot bracket form of the DNA to submit to different softwares if necessary.

      5_3_vector

      5’AACTCCTAAGCCAGTGCCAGAAGAGCCAAGGACAGGT3’ Secondary Structure on Vector Build

      3_5_vector

      3’ TTGAGGATTCGGTCACGGTCTTCTCGGTTCCTGTCCA5’ Secondary Structure on Vector Build

    SimRNA

      This software requires a sequence and secondary structure dot bracket input to give an energy minimized pdb of a secondary structure. However, as the name implies, it only does this for RNA, not DNA. If a DNA sequence is inputted, it converts it to RNA. If we find no other options to get a pdb, along with the options listed previously, we could move forward with SimRNA and assume that the RNA will have a similar binding affinity with ferritin as DNA does.This is one of the assumptions listed at the top of the page in case we move forward with this approach.

You may ask, why are we still doing computational docking if we’ve moved onto the wet lab portion?

Since starting computational docking over the Summer, we’ve also ordered the peptide and 5’-3’ DNA sequence with a biotinylated N terminal and 3’ end, respectively, to be tested in the biotin-streptavidin direct ELISA and DNA shift assay (only DNA). So a question could be, “since we’ve moved on to physical testing, why are we still doing comp dock?” The answer can be broken into a few parts:

    1. A lot of times, the results we get in the lab can be clear but unexplainable. For example, maybe we’re seeing lower or higher than expected binding rates. Computational docking is a great tool to project and make more educated guesses on why this might be happening. This can allow us to better and more quickly target the issue and solve it rather than wasting time and money doing multiple physical experiments to find the problem.

    2. We are also really hoping for the DNA aptamer to work as this would introduce the opportunity of using FRET (Fluorescence Resonance Energy Transfer) for our biosensor. This would be one of the most straightforward and cheap options for our biosensor, allowing us to make it more accessible and usable. Computational docking would allow us to use FRET by allowing us to see the different ways in which our DNA aptamer changes shape to bind to ferritin. This can help us determine best locations on the aptamer for fluorescent tags to be placed. These locations would be determined by seeing what parts of the DNA are close enough to each other in the binding conformation for a fluorescent signal to be released if a pair of tags were placed there. The option to use FRET is definitely one of the reasons why we’re really hoping the DNA aptamer has some strong data from our tests supporting its binding with ferritin. We also expect the DNA to be a bit more flexible and stable than the peptide, making it a better option for a quick and easily storable biosensor.

    • Here are the issues we are currently facing with this:
      • Obviously, as mentioned above, we are struggling to get a pdb or sdf of our DNA’s secondary structure to use in the computational docking software
      • We would need to do a lot more extensive modeling by changing out different bp’s and rerunning the simulations to see which sequences work the best. This would be a much more long-term research model and not something easily done in one year, let alone a semester.

Pdb Files Pulled From the Protein Data Bank:

In order to test the binding affinities of our aptamers through computational docking, we used pdb files of heavy chain and light chain ferritin. Ideally, we would have used a full ferritin pdb. There weren’t any readily available, however, and it would have taken quite a while to create one. Instead, we made the assumption that sterics would not play too big of a role in limiting the aptamers’ binding with ferritin if there were multiple chains instead of just one. Or rather, we assumed that the aptamer having multiple chain options to bind to would counteract the aptamer having issues reaching some of the chains. The final pdb files we ended up using were 2FHA (Human Heavy chain ferritin) and 2FFX (Human Light Chain Ferritin). Below are images of the two chains in PyMol.

2fha

Human Heavy Chain Ferritin (2FHA) in Pymol

2ffx

Human Light Chain Ferritin (2FFX) in Pymol

Citations and References:

Bank, R. P. D. (n.d.). RCSB PDB - 2FHA: HUMAN H CHAIN FERRITIN. https://www.rcsb.org/structure/2fha

Bank, R. P. D. (n.d.-a). RCSB PDB - 2FFX: Structure of Human Ferritin L. Chain. https://www.rcsb.org/structure/2FFX

De Vries, S. J., Van Dijk, M., & Bonvin, A. M. J. J. (2010). The HADDOCK web server for data-driven biomolecular docking. Nature Protocols, 5(5), 883–897. https://doi.org/10.1038/nprot.2010.32

Eberhardt, J., Santos-Martins, D., Tillack, A.F., Forli, S. (2021). AutoDock Vina 1.2.0: New Docking Methods, Expanded Force Field, and Python Bindings. Journal of Chemical Information and Modeling.

J.A. Lemkul (2018) "From Proteins to Perturbed Hamiltonians: A Suite of Tutorials for the GROMACS-2018 Molecular Simulation Package, v1.0" Living J. Comp. Mol. Sci. 1 (1): 5068.

The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC.

SimRNA: a coarse-grained method for RNA folding simulations and 3D structure prediction. Michał J. Boniecki, Grzegorz Łach, Konrad Tomala, Wayne Dawson, Paweł Łukasz, Tomasz Sołtysiński, Kristian M. Rother, and Janusz M. Bujnicki. Nucleic Acids Res. 2015 Dec 19. doi: 10.1093/nar/gkv1479