greenGM/Nbaffinity

This is a nanobody affinity classification tool based on non-covalent interactions of single chain protein in a protein complex.

Nbaffinity: a machine learning tool to predict nanobody affinity

This is a nanoantibody affinity prediction tool based on a machine learning approach that relies mainly on non-covalent interaction data of single chain protein in a protein complex.

GitHub Repository for https://github.com/greenGM/Nbaffinity

You can use this tool directly at： http://www.peptide-ligand.cn/en/

Dependencies:

Name	version
R	4.3.3 or higher
tidyverse	2.0.0 or higher
caret	6.0-94 or higher

data genereation

1. the data was generated by using ProtInter：

   https://github.com/maxibor/protinter

2. All descriptors needs to be generated and fed into the model in the following order:

   HI_ std, HI_ min, HI_ 0.5, HBMS_ count, HBMS_ mean, HBMS_ std, HBMS_ min, HBMS_ max, ASI_ count, ASI_ std, ASI_ min, HBSS_ count, HBSS_ std, HBSS_ min, DB_ count, DB_ std, DB_ 0.5, IoInt_ count, IoInt_ std, IoInt_ min, IoInt_ 0.25, HBMM_ count, HBMM_ mean, HBMM_ std, HBMM_ min, HBMM_ 0.25, HBMM_ 0.5, HBMM_ 0.75, HBMM_ max, CPI_ count, CPI_ std, CPI_ min, AAI_ count, AAI_ std, AAI_ min.

   Abbreviation	Full name
   HI	hydrophobic interactions
   HBMS	hydrogen bonding main-side chain interactions
   ASI	aromatic-sulphur interactions
   HBSS	hydrogen bonding side-side chain interactions
   DB	disulphide bridges
   IoInt	ionic interactions
   HBMM	hydrogen bonding main-main chain interactions
   CPI	cation-pi interactions
   ASI	aromatic-aromatic interactions

How to use this tool:

1. Download all files.

2. Open in Rstudio.

3. Enter the following code：

   R

   load(".RData")

   library(tidyverse)

   library(caret)

   unknown <- read.csv("yourfile.csv",header = F,col.names = name)#Import your data; better without column name, if you have, please set "header = T".

   (Note: the first column of your data should start from "HI_ std" score, not be the protein's name.)

   unknownS <- predict(preprocessParams, unknown)

   unknownprediction <- data.frame(name=unknown$youproteinname, #the name of your protein, if you have.

                              preclass=predict(TdataSW.roF1,unknownS),#Give the class of affinity of the candidate nanobody
   
                              preprob=predict(TdataSW.roF1,unknownS,type = 'prob' ))#Give the probability of the class
   

   write.csv(unknownprediction,'unknownprediction.csv')

   (Note: Please adjust the above code to suit your needs.)

Result explanation：

The result will be returned as a csv file.

 name--the name of the candidate nanobody.

 preclass--predicted class: Y represents MIC < 2000 and N represents MIC ≥ 2000.

 preprob.N preprob.--the probability of the class.