Residue contact

Protein residue contacts are crucial for protein structural prediction and drug target interaction prediction. TMKit allows users to parse protein residue contacts predicted by PSICOV[1], FreeContact[2], CCMPred[3], Gremlin[4], GDCA[5], PlmDCA[6], MemConP[7], Membrain2[8], and DeepHelicon[9].

PSICOV, FreeContact, CCMPred, Gremlin, GDCA, and PlmDCA are canonical covariance methods, while MemConP, Membrain2, and DeepHelicon are machine learning methods specialized for transmembrane proteins.

Reminder of data

Please make sure that the build-in example dataset has been downloaded before you walk through the tutorial.

Example usage

We can use the following code to obtain the residue contacts of protein 1xqf chain A. Similar to other cases in our tutorial, there are commonly used parameters. Please the next section for details.

import tmkit as tmk

df1, df2 = tmk.rrc.read(
    prot_name='1xqf',
    seq_chain='A',
    fasta_fp='data/fasta/',
    pdb_fp='data/pdb/',
    xml_fp='data/xml/',
    dist_fp='data/rrc/',
    tool_fp='data/rrc/tool/',
    seq_sep_inferior=1,
    seq_sep_superior=None,
    tool='membrain2',
)

Attributes

Attribute	Description
pdb_fp	path where a target PDB file is placed
fasta_fp	path where a target Fasta file is placed
xml_fp	path where a target XML file is placed
dist_fp	path where a file containing real distances between residues is placed (please check the file at ./data/rrc in the example dataset)
tool_fp	path where a protein residue contact map file is placed
tool	name of a contact prediction tool. It can be one of PSICOV, FreeContact, CCMPred, Gremlin, GDCA, PlmDCA, MemConP, Membrain2, and DeepHelicon
seq_sep_inferior	The lower bounds of how far any two residues are in pairs
seq_sep_superior	The upper bounds of how far any two residues are in pairs
prot_name	name of a protein in the prefix of a PDB file name (e.g., 1xqf in 1xqfA.pdb)
seq_chain	chain of a protein in the prefix of a PDB file name (e.g., A in 1xqfA.pdb). Parameter file_chain will be converted within the function

See also

Please see here for better understanding the file-naming system.

Output

There are two Pandas dataframes. The first one df1 is the dataframe containing the predicted contacts by tool Membrain2.

print(df1)
contact_id_1  contact_id_2     score
0                13            44  0.032846
1                13            45  0.011669
2                13            46  0.019312
3                13            47  0.089862
4                13            48  0.026575
...             ...           ...       ...
19443           308           349  0.044726
19444           308           350  0.080527
19445           308           351  0.039438
19446           308           352  0.034000
19447           308           353  0.074005
[19448 rows x 3 columns]

The second one df2 is the dataframe containing the real distances between two residues, such that.

Attribute	Description
fasta_id_1	Fasta id of the first residue
aa_1	Amino acid type of the first residue
pdb_id_1	PDB id of the first residue
fasta_id_2	Fasta id of the second residue
aa_2	Amino acid type of the second residue
pdb_id_2	PDB id of the second residue
dist	distance
is_contact	if they are in contact

print(df2)
fasta_id_1 aa_1 pdb_id_1 fasta_id_2 aa_2 pdb_id_2       dist is_contact
0             13    I       15         44    T       46  23.495386          0
1             13    I       15         45    Q       47  22.651615          0
2             13    I       15         46    V       48   18.67347          0
3             13    I       15         47    T       49  19.484049          0
4             13    I       15         48    V       50   21.53894          0
...          ...  ...      ...        ...  ...      ...        ...        ...
19443        308    F      332        349    G      373  35.690994          0
19444        308    F      332        350    Y      374  32.043457          0
19445        308    F      332        351    K      375  38.532841          0
19446        308    F      332        352    L      376  40.355228          0
19447        308    F      332        353    A      377  40.803558          0
[19448 rows x 8 columns]

You can combine the two dataframes directly because they have been aligned this way below, which makes your research easier.

import pandas as pd
df = pd.concat([df1, df2], axis=1)
print(df)

It outputs:

       contact_id_1  contact_id_2     score  ... pdb_id_2       dist is_contact
0                13            44  0.032846  ...       46  23.495386          0
1                13            45  0.011669  ...       47  22.651615          0
2                13            46  0.019312  ...       48   18.67347          0
3                13            47  0.089862  ...       49  19.484049          0
4                13            48  0.026575  ...       50   21.53894          0
...             ...           ...       ...  ...      ...        ...        ...
19443           308           349  0.044726  ...      373  35.690994          0
19444           308           350  0.080527  ...      374  32.043457          0
19445           308           351  0.039438  ...      375  38.532841          0
19446           308           352  0.034000  ...      376  40.355228          0
19447           308           353  0.074005  ...      377  40.803558          0

[19448 rows x 11 columns]

---

[1]

David T. Jones, Daniel W. A. Buchan, Domenico Cozzetto, Massimiliano Pontil, PSICOV: precise structural contact prediction using sparse inverse covariance estimation on large multiple sequence alignments, Bioinformatics, Volume 28, Issue 2, January 2012, Pages 184–190, https://doi.org/10.1093/bioinformatics/btr638

[2]

Kaján, L., Hopf, T.A., Kalaš, M. et al. FreeContact: fast and free software for protein contact prediction from residue co-evolution. BMC Bioinformatics 15, 85 (2014). https://doi.org/10.1186/1471-2105-15-85

[3]

Stefan Seemayer, Markus Gruber, Johannes Söding, CCMpred—fast and precise prediction of protein residue–residue contacts from correlated mutations, Bioinformatics, Volume 30, Issue 21, November 2014, Pages 3128–3130, https://doi.org/10.1093/bioinformatics/btu500

[4]

Sergey OvchinnikovLisa KinchHahnbeom ParkYuxing LiaoJimin PeiDavid E KimHetunandan KamisettyNick V GrishinDavid Baker (2015) Large-scale determination of previously unsolved protein structures using evolutionary information eLife 4:e09248.

[5]

Baldassi C, Zamparo M, Feinauer C, Procaccini A, Zecchina R, et al. (2014) Fast and Accurate Multivariate Gaussian Modeling of Protein Families: Predicting Residue Contacts and Protein-Interaction Partners. PLOS ONE 9(3): e92721. https://doi.org/10.1371/journal.pone.0092721

[6]

Ekeberg M, Lövkvist C, Lan Y, Weigt M, Aurell E. Improved contact prediction in proteins: using pseudolikelihoods to infer Potts models. Phys Rev E Stat Nonlin Soft Matter Phys. 2013 Jan;87(1):012707. doi: 10.1103/PhysRevE.87.012707. Epub 2013 Jan 11. PMID: 23410359.

[7]

Hönigschmid P, Frishman D. Accurate prediction of helix interactions and residue contacts in membrane proteins. J Struct Biol. 2016 Apr;194(1):112-23. doi: 10.1016/j.jsb.2016.02.005. Epub 2016 Feb 3. PMID: 26851352.

[8]

Jing Yang, Hong-Bin Shen, MemBrain-contact 2.0: a new two-stage machine learning model for the prediction enhancement of transmembrane protein residue contacts in the full chain, Bioinformatics, Volume 34, Issue 2, January 2018, Pages 230–238, https://doi.org/10.1093/bioinformatics/btx593

[9]

Sun J, Frishman D. DeepHelicon: Accurate prediction of inter-helical residue contacts in transmembrane proteins by residual neural networks. J Struct Biol. 2020 Oct 1;212(1):107574. doi: 10.1016/j.jsb.2020.107574. Epub 2020 Jul 11. PMID: 32663598.