############################################
#Drugbank XML file extraction (R script)
#Date:         20/01/2016
# Contact:      ashenafi.legehar@helsinki.fi, leo.ghemtio@helsinki.fi
###########################################


#changing the working directory 
setwd('/home/ashenafi/Documents/project/drugdata/DRUGBANK')
#Install XML package in R 
install.packages("XML")
# import XML packge to R enviroment
library(XML)
#the data stored int variable and with internal node 
xml_data1<-xmlTreeParse('drugbank.xml', options=HUGE)
#### data understanding before data mining and extraction###
class(xml_data1)
#Root the xml file 
xmltop = xmlRoot(xml_data1)
###node name
xmlName(c1)
#number of children 
xmlSize(c1)
  # classify the data and store their indices
for (i in 1:length(xml_list))
	{
    #return 1/True if the attribute type is small molecule 
    index=which(xml_list[[i]][['.attrs']][['type']]=='small molecule')
    index2=which(xml_list[[i]][['.attrs']][['type']]=='biotech')
    if (length(index)==1)
    {
      check[i]<-i
    }else{
      biotech[i]<-i
    } 
  }
  #remove all the NA 
  small_index<-check[!is.na(check)]
  biotech_index<-biotech[!is.na(biotech)]
  #######Befor execute the for loop intialize the variables below 
  #index for small molecules
  counter<-1
  #remove all small molecule list
  remove(small_molecule)
  #create all small molecule list 
  small_molecule<-list()
  ###extract all small molelcule######
  for (i in small_index)
  {
  small_molecule[[counter]]<-xml_list[[i]]  
  counter<-counter+1
  }
###extract biotech molecule######
#index for biotech
c<-1
#remove all small molecule list
remove(biotech_m)
#create all small molecule list 
biotech_m<-list()
#getting all biotech molecule in the list 
for (i in biotech_index)
{
  biotech_m[[c]]<-xml_list[[i]]  
  c<-c+1
}
# data extraction selected calculated-properties from drugbank database 

for (i in 1:length(biotech_m))
{
  #drug id                                                                                
  t_a_b[i]<-small_molecule[[i]][[1]][['text']]
  #name of the drug
  t_n_b[i]<-small_molecule[[i]][['name']]
  t_m_b[i]<-0
  #group
  t_b_b[i]<-small_molecule[[i]][['groups']][['group']]
    #Melting Point
  t_m_b[i]<-biotech_m[[i]][['experimental-properties']][[2,1]]
    #Hydrophobicity
  t_h_b[i]<-biotech_m[[i]][['experimental-properties']][[2,1]]
  #Isoelectric Point
  t_i_b[i]<-biotech_m[[i]][['experimental-properties']][[2,2]]
    #Molecular Weight
  t_mw_b[i]<-biotech_m[[i]][['experimental-properties']][[2,3]]
    #Molecular Formula
  t_m_f[i]<-biotech_m[[i]][['experimental-properties']][[2,4]]
}
data_bio<-data.frame(t_a_b, t_n_b, t_b_b), t_m_b, t_h_b, t_i_b, t_mw_b, t_m_f)
View(data_bio)
colnames(data_bio)<-c('id','drug_name','group'), 'Melting Point',
'Hydrophobicity', 'Isoelectric Point', 'Molecular Weight', 
                      'Molecular Formula')
# data extraction selected calculated-properties from drugbank database  
  for (i in 1:length(small_molecule))
  {
    # Initialization the varivable
    t_so[i]<-0
    t_so_t[i]<-'Null'
    l_p[i]<-0
    l_p_s[i]<-'Null'
    t_m[i]<-0
    t_m_s[i]<-'Null'
    t_mo[i]<-0
    t_mo_s[i]<-'Null'
    t_p[i]<-0
    t_p_s[i]<-'Null'
    t_r[i]<-0
    t_r_s[i]<-'Null'
    t_po[i]<-0
    t_po_s[i]<-'Null'
    t_ro[i]<-0
    t_ro[i]<-'Null'
    t_h[i]<-0
    t_h_s[i]<-'Null'
    t_hc[i]<-0
    t_hc_s[i]<-'Null'
    t_pka[i]<-0
    t_pka[i]<-'Null'
    t_pkb[i]<-0
    t_pkb_s[i]<-'Null'
    t_py[i]<-0
    t_py_s[i]<-'Null'
    t_nr[i]<-0
    t_so_t[i]<-'Null'
    t_nr_s[i]<-'Null'
    t_e[i]<-0
    t_f[i]<-'Null'
    #drug id                                                                                
    t_a[i]<-small_molecule[[i]][[1]][['text']]
    name of the drug
    t_n[i]<-small_molecule[[i]][['name']]
    #group
    t_b[i]<-small_molecule[[i]][['groups']][['group']]
    #absorption
    if (is.null(small_molecule[[i]]['absorption'][[1]]))
	{
          t_ad[i]<-'NULL'
    	}else{
      t_ad[i]<-gsub("<.*?>","",small_molecule[[i]]
                    ['absorption'][[1]])
   		 }
    #volume-of-distribution
    if (is.null(small_molecule[[i]]
	{	
                ['volume-of-distribution'][[1]]))
    	{
      t_di[i]<-'NULL'
    }else{
      t_di[i]<-gsub("<.*?>","",small_molecule[[i]]
                    ['volume-of-distribution'][[1]])
    }
    #Metabolism
    if (is.null(small_molecule[[i]]['metabolism'][[1]]))
    {
      t_me[i]<-'NULL'
    }else{
      t_me[i]<-gsub("<.*?>","",small_molecule[[i]]
                    ['metabolism'][[1]])
    }
    #route-of-elimination
    if (is.null(small_molecule[[i]]
                ['route-of-elimination'][[1]]))
    {
      t_el[i]<-'NULL'
    }
    else{
      t_el[i]<-gsub("<.*?>","",small_molecule[[i]]
                    ['route-of-elimination'][[1]])
    }
    #toxicity
    if (is.null(small_molecule[[i]]['toxicity'][[1]]))
    {
       t_to[i]<-'NULL'
    }else{
      t_to[i]<-gsub("<.*?>","",small_molecule[[i]]
                    ['toxicity'][[1]])
    }
    #logP ALOGPS
    if (is.null(small_molecule[[i]]
                [['calculated-properties']][[2,1]]))
    {
      t_c[i]<-0
    }  else
    {
      t_c[i]<-small_molecule[[i]]
      [['calculated-properties']][[2,1]]
    }
    #logP ALOGPS Source
    if (is.null(small_molecule[[i]]
                [['calculated-properties']][[3,1]]))
    {
      t_d[i]<-'Null'
    }else{
      #source 
      t_d[i]<-small_molecule[[i]]
      [['calculated-properties']][[3,1]]
    }
    if(i!=7362 )
    {
    # log S
    if (is.null(small_molecule[[i]]
                [['calculated-properties']][[2,2]]))
    {
      t_so[i]<-0
    } else{
      #logS
      t_so[i]<-small_molecule[[i]]
      [['calculated-properties']][[2,2]]####NewADDED
    }
    #logS ALOGPS
      if (is.null(small_molecule[[i]]
                  [['calculated-properties']][[3,2]] ))
      {
        t_so_t[i]<-'Null'
      }else{
      
        t_so_t[i]<-small_molecule[[i]]
        [['calculated-properties']][[3,2]]####NewADDED
        t_so_t[116]<-'Null'
        t_so_t[205]<-'Null'
      }
      
    	}
    #Water Solublity 
    if(i!=5292 & i!=7362)
    {
      
      if (is.null(small_molecule[[i]]
                  [['calculated-properties']][[2,3]]))
      {
        t_e[i]<-0
      }else{
        #Water Solublity 
        t_e[i]<-small_molecule[[i]]
        [['calculated-properties']][[2,3]]
        t_e[116]<-'Null'
        t_e[205]<-'Null'
      }
      #Wate Solublity source
      if (is.null(small_molecule[[i]]
                  [['calculated-properties']][[3,3]]))
      {
        t_f[i]<-'Null'
      }else {
        #Wate Solublity source 
        t_f[i]<-small_molecule[[i]]
        [['calculated-properties']][[3,3]]
        t_f[116]<-'Null'
        t_f[205]<-'Null'
      }
    }
    #logP ChemAxon, Molecular Weight, Molecular Fromula 
    if(i!=5292 & i!=7362)
    {
      #logP ChemAxon
        if (is.null(small_molecule[[i]
                                   ][['calculated-properties']][[2,4]]))
        {
          l_p[i]<-0
        }else{
          l_p[i]<-small_molecule[[i]]
          [['calculated-properties']][[2,4]] ####newAdded
          l_p[116]<-0
          l_p[205]<-0
        }
        #logP ChemAxon Source
        if (is.null(small_molecule[[i]]
                    [['calculated-properties']][[3,4]]))
        {
          l_p_s[i]<-'Null'
        } else{
          l_p_s[i]<-small_molecule[[i]]
          [['calculated-properties']][[3,4]]
          l_p_s[116]<-'Null'
          l_p_s[205]<-'Null'
        }
          #Molecular Weight
          if (i==116 & i==205)
          {
            t_m[116]<-small_molecule[[i]]
            [['calculated-properties']][[2,4]]
            t_m[205]<-small_molecule[[i]]
            [['calculated-properties']][[2,4]]
          }else if (is.null(small_molecule[[i]]
                           [['calculated-properties']][[2,7]]))
          {
          
              t_m[i]<-0
          }else{
            t_m[i]<-small_molecule[[i]]
            [['calculated-properties']][[2,7]]
          }
        #Molecular Weight source
        if(i==116 & i==205)
        {
          t_m_s[116]<-small_molecule[[i]]
          [['calculated-properties']][[3,4]]
          t_m_s[205]<-small_molecule[[i]]
          [['calculated-properties']][[3,4]]
        }else if (is.null(small_molecule[[i]]
                         [['calculated-properties']][[3,7]]))
        {
          t_m_s[i]<-'Null'
        } else{ 
            t_m_s[i]<-small_molecule[[i]]
            [['calculated-properties']][[3,7]]
        }
        #Molecular Fromula 
        if(i==116 & i == 205)
        {
          t_mo[116]<-small_molecule[[i]]
          [['calculated-properties']][[2,7]]
          t_mo[205]<-small_molecule[[i]]
          [['calculated-properties']][[2,7]]
        }else if(is.null(small_molecule[[i]]
                        [['calculated-properties']][[2,10]]))
        {
          
          t_mo[i]<-0
        }else
        {
            t_mo[i]<-small_molecule[[i]]
            [['calculated-properties']][[2,10]] 
        }
        
        if(i==116 & i==205)
        {
          t_mo_s[116]<-small_molecule[[i]]
          [['calculated-properties']][[3,7]]
          t_mo_s[205]<-small_molecule[[i]]
          [['calculated-properties']][[3,7]]
        }
        #Molecular Formula Source
        else if(is.null(small_molecule[[i]]
                        [['calculated-properties']][[3,10]]))
        {
          t_mo_s[i]<-'Null'
        } else{
            t_mo_s[i]<-small_molecule[[i]]
            [['calculated-properties']][[3,10]]
        }     
    }
    # Polar Surface Area (PSA) Source, Refractivity, Polarizablity, Rotatable_bond_Count
    #H Bond Acceptor Count, H Bond Donor Count,#pKa (strongest acidic) 
    #pKa (strongest basic), Physiological Charge
    if (i!=783 & i!=5292 &i!=7362)
    {
          #Polar Surface Area (PSA) Source 
      if(i==116 & i==205)
      {
        t_p[116]<-small_molecule[[i]]
        [['calculated-properties']][[2,10]]
        t_p[205]<-small_molecule[[i]]
        [['calculated-properties']][[2,10]]
      }else if(is.null(small_molecule[[i]]
                      [['calculated-properties']][[2,13]]))
      {
        t_p[i]<-0
      } else 
      {
      #polar surface area
        t_p[i]<-small_molecule[[i]]
        [['calculated-properties']][[2,13]]

      }
          #Polar Surface Area (PSA) Source 
      if(i==116 & i==205)
      {
        t_p_s[116]<-small_molecule[[i]
                                   ][['calculated-properties']][[3,10]]
        t_p_s[205]<-small_molecule[[i]]
        [['calculated-properties']][[3,10]]
      } else if(is.null(small_molecule[[i]]
                      [['calculated-properties']][[3,13]]))
      {
        t_p_s[i]<-'Null'
      }else 
      {
          t_p_s[i]<-small_molecule[[i]]
          [['calculated-properties']][[3,13]]
      }
        
      if(i==116 & i==205)
      {
        t_r[116]<-small_molecule[[i]]
        [['calculated-properties']][[2,11]]
        t_r[205]<-small_molecule[[i]]
        [['calculated-properties']][[2,11]]
      }
          # Refractivity
      else if (is.null(small_molecule[[i]]
                       [['calculated-properties']][[2,14]]))
      {
        t_r[i]<-0
      }else
      {
          t_r[i]<-small_molecule[[i]]
          [['calculated-properties']][[2,14]] ####newAdded
      }
          #Refractivity Source
      if(i==116 & i==205)
      {
        t_r_s[116]<-small_molecule[[i]]
        [['calculated-properties']][[2,11]]
        t_r_s[205]<-small_molecule[[i]]
        [['calculated-properties']][[2,11]]
      }else if (is.null(small_molecule[[i]]
                       [['calculated-properties']][[3,14]]))
      {
        t_r_s[i]<-'Null'
      } else
      {
          t_r_s[i]<-small_molecule[[i]]
          [['calculated-properties']][[3,14]] ####newAdded
      }
          
      if(i!=595)
      {
              #Polarizablity
        if(i==116 & i==205)
        {
          t_po[116]<-small_molecule[[i]]
          [['calculated-properties']][[2,12]]
          t_po[205]<-small_molecule[[i]]
          [['calculated-properties']][[2,12]]
        }
            
        else if(is.null(small_molecule[[i]]
                        [['calculated-properties']][[2,15]]))
        {
          
          t_po[i]<-0
        }else
        {
          #polarizabiliy
            t_po[i]<-small_molecule[[i]]
            [['calculated-properties']][[2,15]]
        }       
      #polarizablity source 
      if(i==116 & i==205)
      {
        t_po_s[116]<-small_molecule[[i]]
        [['calculated-properties']][[2,12]]
        t_po_s[205]<-small_molecule[[i]]
        [['calculated-properties']][[2,12]]
      }else if(is.null(small_molecule[[i]]
                      [['calculated-properties']][[3,15]]))
      {
        t_po_s[i]<-'Null'
      } else 
      {
          t_po_s[i]<-small_molecule[[i]]
          [['calculated-properties']][[3,15]]
      }    
       if (i!=8 & i!=91)
       {
                # Rotatable_bond_Count
         if(i==116 & i==205)
         {
           t_ro[116]<-small_molecule[[i]]
           [['calculated-properties']][[2,13]]
           t_ro[205]<-small_molecule[[i]]
           [['calculated-properties']][[2,13]]
         }else if(is.null(small_molecule[[i]]
                        [['calculated-properties']][[2,15]]))
          {
            t_ro[i]<-0
          }else
          {
                  #Rotatable_bond_Count 
          t_ro[i]<-small_molecule[[i]]
          [['calculated-properties']][[2,16]]
          }  
          # Rotatable_bond_Count source 
        if(i==116 & i==205)
        {
          t_ro_s[116]<-small_molecule[[i]]
          [['calculated-properties']][[2,13]]
          t_ro_s[205]<-small_molecule[[i]]
          [['calculated-properties']][[2,13]]
        }else if(is.null(small_molecule[[i]]
                        [['calculated-properties']][[3,16]]))
        { 
          t_ro_s[i]<-'Null'
        }else 
        {       
          t_ro_s[i]<-small_molecule[[i]]
          [['calculated-properties']][[3,16]]
        }     
        #H Bond Acceptor Count, H Bond Donor Count,#pKa (strongest acidic) #pKa (strongest basic), Physiological Charge
        if (i!=2428 & i!=3835 &i!=4090 &i!=4979)
          {   
                  #H Bond Acceptor Count
            if(i==116 & i==205)
              {
                    t_h[116]<-small_molecule[[i]]
                    [['calculated-properties']][[2,14]]
                    t_h[205]<-small_molecule[[i]]
                    [['calculated-properties']][[2,14]]
              } else if(is.null(small_molecule[[i]]
                              [['calculated-properties']][[2,17]]))
              { 
                  t_h[i]<-0
              }else
              {
                 #H Bond Acceptor Count             
                  t_h[i]<-small_molecule[[i]]
                  [['calculated-properties']][[2,17]]
                }
                  #H Bond Acceptor Count source 
            if(i==116 & i==205)
            {
              t_h_s[116]<-small_molecule[[i]]
              [['calculated-properties']][[2,14]]
              t_h_s[205]<-small_molecule[[i]]
              [['calculated-properties']][[2,14]]
            }else if(is.null(small_molecule[[i]]
                            [['calculated-properties']][[3,17]]))
            {
              t_h_s[i]<-'Null'
            } else 
                  
           
            {
                #H Bond Acceptor Count source 
              t_h_s[i]<-small_molecule[[i]]
              [['calculated-properties']][[3,17]]
              }
                  
                  
                  if(i!=3231 & i!=4114 & i!=4142 & i!=4244 )
                  {
                    
                    #H Bond Donor Count
                    if(is.null(small_molecule[[i]]
                               [['calculated-properties']][[2,17]]))
                    {
                      
                      t_hc[i]<-0
                    }else
                    {
                      #H Bond Donor Count
                      if(i==116 & i==205)
                      {
                        t_hc[116]<-small_molecule[[i]]
                        [['calculated-properties']][[2,15]]
                        t_hc[205]<-small_molecule[[i]]
                        [['calculated-properties']][[2,15]]
                      }else
                      {
                        t_hc[i]<-small_molecule[[i]]
                        [['calculated-properties']][[2,18]]
                      }
                    }
                    
                    if(is.null(small_molecule[[i]]
                               [['calculated-properties']][[3,18]]))
                    {
                      t_hc_s[i]<-'Null'
                    }else 
                    {
                      #H Bond Donor Count source
                      if(i==116 & i==205)
                      {
                        t_hc_s[116]<-small_molecule[[i]]
                        [['calculated-properties']][[2,15]]
                      } else
                      {
                        t_hc_s[i]<-small_molecule[[i]]
                        [['calculated-properties']][[3,18]]
                      }
                    }
                  
                    #Pka
                    if(i!=1811 & i!=2683 &i!=3592)
                    {
                    
                      if(is.null(small_molecule[[i]]
                                 [['calculated-properties']][[2,19]]) )
                      {
                        
                        t_pka[i]<-0
                      }else
                      {
                        #pKa (strongest acidic)
                        if(i==116 & i==205)
                        {
                          t_pka[116]<-small_molecule[[i]]
                          [['calculated-properties']][[2,16]]
                          t_pka[205]<-small_molecule[[i]][['calculated-properties']][[2,16]]
                        } else
                        {
                          t_pka[i]<-small_molecule[[i]]
                          [['calculated-properties']][[2,19]]
                        }
                      }
                      #pKa (strongest acidic) source
                      
                      if(is.null(small_molecule[[i]]
                                 [['calculated-properties']][[3,19]] ))
                      {
                        
                        t_pka_s[i]<-'Null'
                        
                      }else 
                      {
                        #pKa (strongest acidic) source
                        if(i==116 & i==205)
                        {
                          t_pka_s[116]<-small_molecule[[i]]
                          [['calculated-properties']][[2,16]]
                          t_pka_s[205]<-small_molecule[[i]]
                          [['calculated-properties']][[2,16]]
                        }else
                        {
                          t_pka_s[i]<-small_molecule[[i]]
                          [['calculated-properties']][[3,19]]
                        }
                      }
                      if(i!=859 & i!=927 & i!=1204 & i!=1215 & i!=1219 & i!=1324 & i!=1415 
                         & i!=1416 & i!=1973 & i!=2903 & i!=3097 & i!=4143 & i!=4637 &
                         i!=4721 & i!=4954 & i!=4968 & i!=5306 & i!=5330 & i!=7396)
                      {
                        if(is.null(small_molecule[[i]]
                                   [['calculated-properties']][[2,20]]))
                        {
                          t_pkb[i]<-0
                        }else
                        {
                          #pKa (strongest basic)
                          if(i==116 & i==205)
                          {
                            t_pkb[116]<-small_molecule[[i]]
                            [['calculated-properties']][[2,17]]
                            t_pkb[205]<-small_molecule[[i]]
                            [['calculated-properties']][[2,17]]
                          }else
                          {
                            t_pkb[i]<-small_molecule[[i]]
                            [['calculated-properties']][[2,20]]
                          }
                        }
                        if(is.null(small_molecule[[i]]
                                   [['calculated-properties']][[3,20]]))
                        {
                          
                          t_pkb_s[i]<-'Null'
                        }else 
                        {
                          #pKa (strongest basic)
                          if(i==116 & i==205)
                          {
                            t_pkb_s[116]<-small_molecule[[i]]
                            [['calculated-properties']][[2,17]]
                            t_pkb_s[205]<-small_molecule[[i]]
                            [['calculated-properties']][[2,17]]
                          } else
                          {
                            t_pkb_s[i]<-small_molecule[[i]]
                            [['calculated-properties']][[3,20]]
                          } 
                        }
                      } 
                    }
                  }                  
                }          
             }
          }  
    }
  } 
#Store the extracted data in to dataframe
data<-data.frame(t_a,t_n,t_b,as.numeric(t_c),t_d,as.numeric(t_so), t_so_t,
                  as.numeric(l_p),l_p_s,as.numeric(t_m),t_m_s,t_mo,t_mo_s, 
                 as.numeric(t_p), t_p_s, as.numeric(t_r), t_r_s,
                  as.numeric(t_po),t_po_s, as.numeric(t_ro), t_ro_s, 
                 as.numeric(t_h), t_h_s,as.numeric(t_hc), t_hc_s, as.numeric(t_pka),
                 t_pka_s, as.numeric(t_pkb), t_pkb_s, as.numeric(t_py), t_py_s, t_ad,
                 t_di,t_me,t_el, t_to)
#Rename the extracted data column
colnames(data)<-c('id','drug_name','group','logP', 'source_logP','logS', 'source_logS',
                   'logP ChemAxon', 'source_logP ','Molecular_weight', 'Mweight_source', 
                   'Molecular_Formula','M_source','Polar_Surface_Area', 'PSA_source', 
                  'Refractivity','Refractivity_source', 'Polarizability', 'Polarizability_source',
                  'Rotatable_bond_Count', 'RBC_source', 'Hydrogen_bond_Acceptor', 'HBAC_source', 
                  'Hydrogen_bond_donor', 'BDC_source','Pka_acidic', 'Pka_source', 'Pka_basic',
                  'Pka_basic_source','Pysiological_charge', 'PC_source', 'Absorption',
                  'Volume-of-distribution','Metabolism',
                  'route-of-elimination','Toxicity')
#export the extracted file as CSV
write.csv(data, "drugbank.csv",row.names=FALSE)