ATP GTP Problem 2 mm.py

Transcription

1 Problem 1 This problem will give you some experience with the Protein Data Bank (PDB), structure analysis, viewing and assessment and will bring up such issues as evolutionary conservation of function, structure and sequence. For a large portion of this problem you will need to use the molecular visualization program PyMOL. PyMOL is freely available for a number of different platforms (see A neat thing about PyMOL is that it has a python interpreter embedded into it, so you can write molecular visualization scripts using python! It also makes attractive images, which you may find useful in your future scientific career (or for class presentations). PyMOL is well documented on the web and you will find that answers to many questions about how to use it can be easily found with google. The official PyMOL User s Manual and Reference Manual can be found at There is also a PyMOL wiki that is a good source of commands and other PyMOL information: By using these resources, and just playing around with the program, you will quickly pick up the small number of commands required to do the following assignment. PyMOL is available on Athena. To use it type add pymol followed by pymol at the Athena prompt. Additionally, you can install it on your home machine. If you have problems using PyMOL, do not hesitate to contact your TAs. Premise for problem 1 You have set out to computationally re-design a novel signal-regulated kinase protein to serve as a convenient read-out for a particular signaling pathway (an ambitious task!). Computational protein design usually begins with the known structure of an existing protein, as you will see later in this course. Among native kinases, you have limited your search for an appropriate template to two candidates: - Human mitogen-activated protein kinase 2 (PDB ID 1S9I) - Rat mitogen-activated protein kinase ERK2 (PDB ID 1ERK) Assessment of Structure Quality a) Go to the Protein Data Bank website ( and navigate to the entries for the two proteins listed above. Several parameters that can be used to assess structure quality have been discussed in class. Some of these are: the resolution, the R value, R free and the average B-factor. For each of these parameters, explain what it reflects, give the relevant value for each structure, and indicate which of the two structures is better in terms of each parameter. Hint: you may find useful information in different places: e.g. on the front page of the PDB entry, in the PDB header itself, or in summary files generated by the PDB under the Structure Analysis:Geometry section on the left panel. b) Using the Procheck tool under Structure Analysis:Geometry, analyze the backbone conformation of each structure using the Ramachandran plot. Which structure has a better backbone geometry? (Give quantitative support for your answer.) Why are Pro and Gly treated specially in this analysis?

2 Structure-based Sequence Alignment c) Go to the Pairwise Comparison of Protein Structures portion of the Dali server ( Note, 1S9I is crystallized as a dimer (chains A and B), so you should try two alignments 1ERK with 1S9I chain A and 1ERK with 1S9I chain B. Enter the PDB codes of the proteins above and specify either chain A or B for 1S9I (in two separate runs) to align the structures. Pick the alignment with the highest Z-score. How many residues were successfully aligned? What is the RMSD in the aligned region? What fraction of amino acids in the alignment are identical between the two proteins? Do not discard the results page of this alignment, you will need it later. d) Download the files mol1_original.pdb and mol2_1.pdb generated by DALI for the alignment of 1ERK with 1S9I chain B. mol2_1.pdb is the second molecule you entered, rotated so that it is aligned with the first. Rename these two objects in PyMOL appropriately as 1S9I and 1ERK by clicking A next to the object and selecting rename (make sure you label them correctly!). Play around with hiding and showing different features (side chains, ribbons, cartoon, etc) as well as with coloring (under C to the right of object names) of the two molecules to assess the quality of the alignment. As you can deduce from its PDB entry, 1S9I has not yet been classified by SCOP. In your opinion, what is the lowest level in SCOP hierarchy that should be shared by 1ERK and 1S9I? e) Go to the Pairwise Alignment Algorithms section of the EMBOSS website ( and use the water tool (Smith-Waterman algorithm) to generate the best local alignment between chain B of 1S9I and 1ERK (get the FASTA sequences from the PDB). Use a BLOSSUM40 matrix, a gap penalty of 20 and leave all other parameters at their default values (empirically, this set of parameters produces good alignments for these proteins). Find the two largest gaps predicted by the Smith-Waterman approach and the two largest gaps assigned by Dali. Locate these gaps in the structures (use the same PyMOL session as in part (d) to simultaneously view the two structures aligned and use the sequence viewer to select different regions). Comment on the differences and similarities of placement for these gaps between Dali and Smith- Waterman. Comment on the appropriateness of the gaps given the structural alignment. Which of the two sequence alignments is more useful? Attach a printout of both alignments with the two gaps in question circled. f) As you may have noticed, 1S9I is co-crystallized with ATP. In the same PyMOL session as in part (d), type select LIG, resn ATP into the PyMOL command line to create a selection called LIG that consists of all atoms belonging to residues named ATP (resn stands for residue name ). How many atoms are selected (see output in command window)? What is the chemical formula for ATP? How many atoms are in ATP in total? Why is this count different from the PyMOL count? g) Type select AS, LIG around 4 in 1S9I into PyMOL command line to create a selection called AS that contains all atoms in chain B of 1S9I within 4 Å of

3 selection LIG. From the menu bar select Display:Sequence. Two rows of sequences (corresponding to the two molecules read) are shown. Residues with selected atoms are highlighted in the sequence for 1S9I. Report the selected residues and their indices (note, sequence indices correspond to the amino acid in the same column as the first digit of the index). h) Go back to the Dali results page and click on the link under the Structural Alignment column in the results table. Using this alignment, calculate the percent identity of residues in your selection AS and report this. Explain the sizeable discrepancy between the number you obtained here and the overall sequence identity of the alignment. Redesigning binding specificity: i) You realize that as part of your reengineering of a native kinase, you may want to change its binding preference from ATP to GTP (see structures below). Use PyMOL to analyze the ATP molecule bound to 1S9I. Out of the residues in selection AS, which 4 would you consider most likely candidates for mutations that change the binding specificity of your kinase. ATP GTP Problem 2 In this problem you will get practice parsing through a PDB file, identifying specific atoms and calculating several molecular mechanics energy terms. a) Write a python script called mm.py that accepts the name of a PDB file as the first command line argument, calculates C -C bond stretching energies for all residues (except glycine, which of course does not have a C atom) and van der Waals energies for all interactions between backbone amide N atoms and backbone carbonyl O atoms occurring between residues separated by at least 1 other residue. Your script should only print those bond stretching and van der Waals energies that are above the cutoff energy provided as the second command line argument. A suggested algorithm for this program is provided below. You can, if you want to, choose to use your own algorithm. Your program is going to

4 be tested based on its output. As before, please make sure you format your output exactly as shown in the sample run below. b) Download the structure of entry 1BBG from the PDB. Run your program on this structure with an energy cutoff of 0.1 kcal/mol. What can you say about the relative magnitudes of van der Waals versus bond-stretching energies? Does this make sense in terms of the equations below? Note: if you can t get your code to work, use the sample run below to answer this question. c) For 1BBG, do you see any pattern of residue index pairs for which van der Waals interactions exceed the cutoff? Is this pattern consistent with any particular structure type? Open 1BBG in PyMOL and examine the interactions that you detected to have van der Waals interactions above 0.1. Why would these atom pairs be closer to one another than the distance at which they have optimal van der Waals interactions (equilibrium distance)? Atom types, formulae and parameters for MM terms As you have seen in class, bond-stretching energies are calculated according to the formula: 1 U ( ) 2 bond = kb b bo, 2 where k b and b o are the force constant and equilibrium length for the given bond and b is the actual bond length. For C -C bonds that you have to consider in this assignment, k b is kcal/å 2 and b o is 1.53 Å. The formula you will use for calculating van der Waals interactions in this assignment is as follows: 12 6 Rmin Rmin U vdw = ε 2, rij rij where and R min are the energy well-depth and equilibrium distance for the given atom pair and r ij is the actual distance (in Angstroms) between the two atoms. For amide N and carbonyl O interactions that you have to consider in this problem, is kcal/mol and R min is 3.2 Å. In PDB files C, C, backbone N and backbone O atoms have atom names of CA, CB, N and O (see PDB file format specifications below). You can rely on this fact in your script (as does the algorithm below). Additionally, you can rely on the fact that if there is a CB atom in a residue, it always comes after a CA atom in that same residue. PDB file format In this problem we will only be concerned with parsing ATOM records of a PDB file (i.e. only lines starting with ATOM ). Each ATOM record is one line and contains information about one atom its name, its index in the structure, the name and the index of the residue it belongs to, its x, y, and z coordinates and its temperature factor, to name a few. Each of these values resides in a specific range of characters within the ATOM record string. For example, atom name is always found between characters 13 and 16

5 (inclusively) and residue name between characters 18 and 20 (inclusively). For the full specification of the ATOM record format, see Keep in mind that the ranges given assume that the first character has an index of 1, whereas in python indexing starts at 0, so in your program you have to adjust your character ranges accordingly. Below is an example ATOM record (in fixed-width font) along with a meter for your convenience. ATOM 14 2HB ASP H Suggested algorithm Algorithms are often communicated in terms of pseudo-code something half way between a plain human language and a computer language. Pseudo-code uses basic constructs that are common to most computer languages (such as variables, assignments, loops and functions) while at the same time abstracting away from any particular language thus allowing for the flexibility to easily explain most important aspects of an algorithm. Below is a suggested algorithm for solving this problem written in pseudocode. Comments are in plain text while code is in bold. process command-line arguments bind pdbf to the input PDB file name bind ecut to the specified energy cutoff Arrays Xn, Yn, and Zn will hold x, y, and z coordinates of backbone amide nitrogen (N) atoms in the order encountered in the structure. Arrays rnn and rin will hold residue names and indices of the corresponding atoms. Thus, Xn[i], Yn[i], Zn[i] will hold the coordinates of the i-th N atom in the structure and rnn[i] and rin[i] will hold the residue name and number of the residue to which this atom belongs. Arrays Xo, Yo, Zo, rno, and rio will hold the same information for backbone carbonyl oxygen (O) atoms. We will need this information for calculating distances and for producing understandable output. initialize empty arrays Xn, Yn, Zn, rnn, and rin initialize empty arrays Xo, Yo, Zo, rno, and rio for each line in file pdbf: if first 4 characters of the line are not ATOM : skip this line Given the specification of the PDB format above, accessing specific values on each line is a simple matter of looking at the right sub-string. However, make sure to use the strip() function (i.e. str = str.strip() ) to remove any trailing or leading white space characters from each value. bind atom name to variable an we are only interested in CA, CB, N and O atoms if atom name is neither of CA, CB, N or O : skip this line bind atom x-coordinate to variable x bind atom y-coordinate to variable y bind atom z-coordinate to variable z bind residue name to variable rn bind residue index to variable ri if we just found a CA atom, we need to remember this atom and update f to indicate that a CA has been seen in the current residue

6 if atom name is CA: remember the information about the 1 st atom in our CA-CB bond x1 = x y1 = y z1 = z rn1 = rn ri1 = ri f = 1 if atom name is CB, that means a previous CA had to have been encountered (because of their order in a PDB file). if atom name is CB: calculate the distance between CA-CB. bind b to the distance between [x1, y1, z1] [x, y, z] bind en to the CA-CB bond-stretching energy at distance b if en is grater than ecut: print Bond %s %3d CA - %s %3d CB %.2e % (rn1, ri1, rn, ri, en) if atom name is either N or O we need to store information about this atom for later calculation of N-O van der Waals energies if atom name is N : append arrays Xn, Yn, and Zn with x, y, and z respectively append arrays rnn and rin with rn ri respectively if atom name is O : append arrays Xo, Yo, and Zo with x, y, and z respectively append arrays rno and rio with rn ri respectively Now we are ready to iterate over all pairs of O-N atoms and calculate each interaction energy for each i-th backbone atom O and j-th backbone atom N: we only care about interactions non-local in sequence not those within the same residue or between neighboring residues use residue index to determine this if the two atoms are in the same residue or adjacent residues: skip this iteration use values in arrays Xn, Yn, Zn, Xo, Yo, Zo to calculate distance d = distance between the two atoms en = van der Waals energy of N-O interaction at distance d if en is grater than ecut: print vdw %s %2d O - %s %2d N %6.2f % (rno[i], rio[i], rnn[j], rin[j], en) Python Hints - In python, sub-strings (or slices of arbitrary arrays) can be easily accessed by specifying subscript ranges. Consider the following code snippet: str = "some long string of characters" print str print str[0:3] print str[12:18] print str[:18] print str[7:] print str[:]

7 This code produces: some long string of characters som ring o some long string o ng string of characters some long string of characters Notice several things: Just like with the range() function, the end index is treated exclusively. That is, if the ending index is 5, the character with index 5 is the first NOT to be included in the substring. Both the beginning and the ending indices are optional the beginning index defaults to 0 and ending index to the length of the string. - If you have read in a value as a string and would like to convert it into a float, use python s built-in function float(). The same works for integers and function int(). Consider the following code: s1 = " " s2 = "11" f = float(s1) i = int(s2) print f, i, f+1, i*i which produces: To calculate the distance between two points in 3D located at [x i, y i, z i ] and [x j, y j, z j ] we can apply the Pythagorean theorem twice to arrive at the formula: d = ( x x ) + ( y y ) 2 + ( z z ) 2 i 2 j i j i j - In order to calculate exponents in python, use the built-in function pow(). pow(x, y) calculates x y. Sample run The run below is done on the PDB entry with ID 1BBG. It is also posted on the class website as sample.run.txt. [gevorg@keating9 ps3]$ python mm.py 1BBG.pdb 0.1 Bond LYS 13 CA - LYS 13 CB 1.46e-01 Bond TYR 29 CA - TYR 29 CB 1.21e-01 Bond ILE 34 CA - ILE 34 CB 3.13e-01 Bond ASN 38 CA - ASN 38 CB 2.14e-01 vdw SER 31 O - ALA 33 N 3.60 vdw SER 31 O - ILE 34 N 27.63

8 vdw LYS 32 O - CYS 35 N 1.13 vdw ALA 33 O - CYS 35 N 0.23 vdw ALA 33 O - LYS 37 N 0.94 vdw ILE 34 O - CYS 39 N 0.26 vdw CYS 35 O - CYS 39 N 0.70 vdw ASN 38 O - THR 40 N 0.10