chemoinformaticsな用途でbayonを使ってみる。

データセットはPrimary screen for compounds that inhibit Insulin promoter activity in TRM-6 cells.でアクティブだった1153検体

ダウンロードしてきたsdfをopenbabelでフィンガープリントに変換

babel -imol pc_sample.sdf -ofpt pc_sample.fpt -xh -xfFP2

これをbayonでクラスタリングにするためのTSVに変換

python f2bayon.py pc_sample.fpt > pc_sample.tsv

f2bayon.pyのソース

def hex2bin(fingerprint):
   bf = ""
   h2b = {"0":"0000","1":"0001","2":"0010","3":"0011",
          "4":"0100","5":"0101","6":"0110","7":"0111",
          "8":"1000","9":"1001","a":"1010","b":"1011",
          "c":"1100","d":"1101","e":"1110","f":"1111",
          }

   for l in fingerprint:
       for c in l:
           b = h2b.get(c)
           if b: bf += b
   return bf

def convert(file):
   result = ""
   for data in open(file,"r").read().split("\n>"):
       fp = ""
       for list in data.split("\n")[1:]:
           fp += hex2bin(list)
       result += data.split("\n")[0].split(" ")[0] + " " + fp + "\n"
   return result

if __name__ == "__main__":
   import sys
   file = sys.argv[1]
   c = convert(file)
   for l in c.split("\n")[:-1]:
      id,fp = l.split()
      fps = ""
      for num,bit in enumerate(fp):
         if int(bit) > 0:
            fps += "\t%d\t%s" % (num,bit)
      print id + fps

でbayonで10クラスターに分割

$ time bayon -n 10 pc_sample.tsv > pc_sample.clust

real    0m0.859s
user    0m0.839s
sys 0m0.015s

超速い。