This project can build a data set of helices.
We saved ten high-resolution PDBs in PDB/standard_set after downloading them from the PDB website. Then we manually checked these 10 PDBs, extracted the good helices (we hope), and saved the residues' name and sequence in standard_helices.xlsx so that we could edit them easily. We used the script read_xls.py to read records from the standard.xlsfile, then adjusted the helix records in the PDB file based on the records read from the xls file we prepared previously, and then saved the new modified PDBs in the manual_set.
We also want to get a training set automacally, so we use train.py modified helices records in PDBs from standard_set and saved the modified PDBs in the train_set.
Finally, we compared the training PDBs (train_set) with the PDBs modified manually (manual_set) by evaluate.py to evaluate the training process (train.py) if it is useful.
Reading standard_helices.xlsx, changing the helices in the standard_set PDBs, and saving the updated PDBs to the manual_set directory.
PDBs in the manual_set are inputed and modified PDBs are outputted to the training_set directory. To make helices for the train set, we follow these rules:
- Getting alpha-helices from PDB files and splitting the helix from Proline.
- Merging the ksdssp, cablam and dsssp annotating results.
- Using phi-psi angles to reduce redundancy residues.
Checking each residue's hydrogen bond length and cutting those that don't match the theoretical bond length(3.5).
comparing PDBs in the manual_set with training_set and evaluatting the result. The result include 5 parameters: helix_ID, Status, length, name and sequeen of start residue, name and sequeen of end residue. For example:
| ID | Status | difference | start_res | end_res |
|---|---|---|---|---|
| 1 | identical | 0 | GLU 59 | ALA 59 |
| 2 | longer | 1 | ALA 22(ASP 20) | HIS 36(SER 35) |
It means that both PDBs have helix 1 and helix 2, but helix 1 from the manual set and the training set are identical, whereas helix 2 which from the manual set and form the training set are similar but have different start and end residues.
There are 6 parameters of Status: more, lost, shorter, longer, identical, same.
| Parameters | Description |
|---|---|
| more | the training PDB has the 'wrong' helix which is not in the manual PDB. |
| lost | the training PDB lost the helix which is in the manual PDB. |
| identical | tow helices are identical. |
| same | the length of two helices are identical but start residues are differnt. |
| shorter | the helix from PDBs in training set is shorter than manual set. |
| longer | the helix from the training set is longer than manual set. |
Difference means length difference of tow helices: length of helix1 - length of helix2.
*The max diffrence of longer and shorter less than 4 reisidues.
Runing them srep by step:
read_xls.py: phenix.python read_xls.py standard_helices.xlsx
train.py: phenix.python train.py PDB/standard_set/
H_bond.py:phenix.python H-bond.py PDB/traning_set/
evaluate.py: phenix.python evaluate.py PDB/training_set PDB/manual_set