Skip to content

Commit

Permalink
revise (#2)
Browse files Browse the repository at this point in the history 
* update input for program: now only requires receptor and poses file
* add test cases
* clean up README / installation notes

Co-authored-by: karoka <xjr,karo@gmail.com>
  • Loading branch information
@karoka
karoka and karoka authored Apr 27, 2020
1 parent ee45bcc commit 866fdd6
Show file tree
Hide file tree
Showing 1,367 changed files with 111,244 additions and 2,032,225 deletions.
91 changes: 45 additions & 46 deletions README.md
View file
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
# RNAPosers
RNAPosers: Machine-Learning Pose Classifiers for RNA Containing Complexes.

This repo contains source code for RNAPosers' pose fingerprint and prediction (classification) modules. Pose fingerprint module is written in C++, and will generate executable `bin/featurize` once compiled. Prediction (classification) is done using Python script `src/rna_poser.py` and pre-trained classifiers with various parameter setting in `classifier/` . All the classifiers were trained with `Python 3.5.3` and `sklearn v0.19.2`, but should be compatible with both Python2 and Python 3 and any later version sklearn. The combined fingerprinting and classification process can be done by running `rna_posers.sh`.
This repo contains source code for RNAPosers' pose fingerprint and prediction (classification) modules. Pose fingerprint module is written in C++, and will generate executable `bin/featurize` once compiled. Prediction (classification) is done using Python script `src/rna_poser.py` and pre-trained classifiers with various parameter setting in `classifier/` . All the classifiers were trained with `Python 3.5.3` and `sklearn v0.19.2`, but should be compatible with both Python2 and Python 3 and any later version sklearn. The combined fingerprinting and classification process can be done by running `src/run.sh`.

**RNAPosers-plugin**: We also offer a PyMOL plugin version of RNAPosers (`rnaposerplugin.zip`) with graphical user interface as a supplement to the source code version. See [Using RNAPosers PyMOL plugin](#Using-RNAPosers-PyMOL-plugin) for installation instructions.

Expand All @@ -23,73 +23,76 @@ source ~/.bashrc

### Setup environment
```
conda create --name rnaposers
conda env create -f env/rnaposers.yml
conda activate rnaposers
conda install -c anaconda scikit-learn
conda install -c conda-forge openbabel
```
<!---
# conda create --name rnaposers
# conda activate rnaposers
# conda install -c schrodinger pymol
# conda install -c schrodinger pymol-psico
# conda install -c openbabel openbabel
# conda install pandas
# conda install -c anaconda scikit-learn
-->

### Using RNAPosers

```
./src/rna_poser.sh -h
usage: ./rna_poser.sh [base-directory]
[pdbid]
[ref-pdb-name]
[ref-mol2-name]
[ref-dcd-name]
[rmsd-threshold: 1., 1.5, 2., or 2.5]
[eta: 2, 24, or 248]
[feature-file-prefix]
[class-scores-prefix]
example: ./rna_poser.sh tests/input/2b57/ 2b57 complex.pdb lig_2b57.mol2 complexes.dcd 2.5 248 output/2b57_features output/2b57_class_scores
./src/run.sh -h
```

#### Arguments
- **base-directory**: the path to folder that contain the receptor-ligand pdb file, ligand mol2 file, receptor-ligand poses dcd file
- **pdbid**: identifier of current structure
- **ref-pdb-name**: name of reference RNA-ligand pdb file]
- **ref-mol2-name**: name of reference ligand mol2 file]
- **ref-dcd-name**: name of receptor-ligand pose dcd file]
- **rmsd-threshold**: definition of nativeness. Either: `1`, `1.5`, `2`, or `2.5`
- **eta**: a Guassian width parameter for pose fingerprint. `eta=2` means {2 Å} (used in the manuscript), `eta=24` means {2Å and 4 Å}, and `eta=248` means {2, 4, and 8 Å}. The higher the eta values, the more complex the fingerprint and longer the time will take for the computation of pose fingerprint.
- **feature-file prefix**: prefix to pose fingerprint file
- **class-score file prefix**: prefix to which classification scores output file
- **receptor mol2**: .mol2 file of receptor structure. Default: tests/input/1AM0/receptor.mol2
- **ligand poses sd**: .sd file containing all ligand posees. Default: tests/input/1AM0/poses.sd
- **output file**: where to save scores. Default: tests/output/1AM0.txt
- **rmsd**: 1, 1.5, 2, 2.5. Default: 2
- **eta**: 2, 24, or 248 (2A, 2A and 4A, 2A 4A and 8A). Default: 248
- **stop frame**: only score the first several poses. Default: -1 (using all frames)

### Example
```
cd $RNAPOSERS_PATH
pdb=2b57
mkdir tests/output/
./src/rna_poser.sh tests/input/${pdb}/ ${pdb} complex.pdb lig_${pdb}.mol2 complexes.dcd 2.5 2 tests/output/${pdb}_features tests/output/${pdb}_class_scores
# Test 1: score the first 10 poses of 1AM0
./src/run.sh tests/input/1AM0/receptor.mol2 tests/input/1AM0/poses.sd tests/output/1AM0.txt 10 2.5 248
# Test 2: score the full set of poses of 2B57
./src/run.sh tests/input/2B57/receptor.mol2 tests/input/2B57/poses.sd tests/output/2B57.txt
```
#### Output
```
cat tests/output/class_scores_${pdb}.txt
cat tests/output/1AM0.txt
# columns: prediction probability(0) probability(1)
1.000000 0.027000 0.973000
1.000000 0.023000 0.977000
1.000000 0.009000 0.991000
1.000000 0.062000 0.938000
1.000000 0.008000 0.992000
1.000000 0.050000 0.950000
1.000000 0.031000 0.969000
1.000000 0.010000 0.990000
1.000000 0.284000 0.716000
0.000000 0.687000 0.313000
cat tests/output/2B57.txt
1.000000 0.025000 0.975000
1.000000 0.015000 0.985000
1.000000 0.025000 0.975000
1.000000 0.005000 0.995000
1.000000 0.004000 0.996000
1.000000 0.001000 0.999000
1.000000 0.004000 0.996000
1.000000 0.002000 0.998000
1.000000 0.001000 0.999000
... ... ...
... ... ...
... ... ...
0.000000 0.963000 0.037000
0.000000 0.967000 0.033000
0.000000 0.975000 0.025000
0.000000 0.983000 0.017000
0.000000 0.984000 0.016000
1.000000 0.004000 0.996000
1.000000 0.005000 0.995000
1.000000 0.005000 0.995000
1.000000 0.004000 0.996000
...
```

### PyMOL plugin
The RNAPosers PyMOL plugin is compressed it as `rnaposerplugin.zip`. See PyMOL website: https://pymolwiki.org/index.php/Plugin_Manager for an instruction on installing PyMOL plugin from local file. Note that you still have to install this repo and setup environment to use the plugin.


### Additional Notes
- RNAPosers expect that the ordering of ligand atoms in the reference receptor-ligand pdb matches the order in the reference ligand mol2 file
- RNAPosers reads atom types from the reference ligand mol2 file
- Atom types should to be one of the following SYBYL atom types:
- Ligand Atom types should to be one of the following SYBYL atom types:

Description | Type
--- | ---
Expand All @@ -115,10 +118,6 @@ Sulphone sulphur | S.o2
Phosphorus sp3 | P.3






## License
```
Copyright University of Michigan.
Expand Down
90 changes: 90 additions & 0 deletions env/rnaposers.yml
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,90 @@
name: rnaposers
channels:
- openbabel
- anaconda
- schrodinger
- defaults
dependencies:
- apbs=1.5=h1de35cc_3
- asn1crypto=1.3.0=py37_0
- blas=1.0=mkl
- bzip2=1.0.8=h1de35cc_0
- ca-certificates=2020.1.1=0
- cairo=1.14.12=hc4e6be7_4
- certifi=2020.4.5.1=py37_0
- cffi=1.14.0=py37hb5b8e2f_0
- chardet=3.0.4=py37_1003
- cryptography=2.8=py37ha12b0ac_0
- dbus=1.13.12=h90a0687_0
- expat=2.2.6=h0a44026_0
- fontconfig=2.13.0=h5d5b041_1
- freemol=1.158=py37_1
- freetype=2.9.1=hb4e5f40_0
- gettext=0.19.8.1=h15daf44_3
- glew=2.0.0=0
- glib=2.63.1=hd977a24_0
- h5py=2.10.0=py37h3134771_0
- hdf5=1.10.4=hfa1e0ec_0
- icu=58.2=h4b95b61_1
- idna=2.9=py_1
- intel-openmp=2019.4=233
- joblib=0.14.1=py_0
- jpeg=9b=he5867d9_2
- libcxx=4.0.1=hcfea43d_1
- libcxxabi=4.0.1=hcfea43d_1
- libedit=3.1.20181209=hb402a30_0
- libffi=3.2.1=h0a44026_6
- libgfortran=3.0.1=2000
- libiconv=1.15=hdd342a3_7
- libopenblas=0.3.6=hdc02c5d_2
- libpng=1.6.37=ha441bb4_0
- libtiff=4.1.0=hcb84e12_0
- libxml2=2.9.9=hf6e021a_1
- llvm-openmp=4.0.1=hcfea43d_1
- mengine=1=h1de35cc_1
- mkl=2019.4=233
- mkl-service=2.3.0=py37hfbe908c_0
- mkl_fft=1.0.15=py37h5e564d8_0
- mkl_random=1.1.0=py37ha771720_0
- mpeg_encode=1=h1de35cc_1
- mtz2ccp4_px=1.0=hdc02c5d_3
- ncurses=6.2=h0a44026_0
- numpy=1.18.1=py37h7241aed_0
- numpy-base=1.18.1=py37h6575580_1
- olefile=0.46=py37_0
- openbabel=2.4.1=py37_6
- openssl=1.1.1g=h1de35cc_0
- pandas=1.0.3=py37h6c726b0_0
- pcre=8.43=h0a44026_0
- pdb2pqr=2.1.1=py37_1
- pillow=7.0.0=py37h4655f20_0
- pip=20.0.2=py37_1
- pixman=0.38.0=h1de35cc_0
- pmw=2.0.1=py37_2
- pycparser=2.20=py_0
- pymol=2.3.5=py37h95b93ae_0
- pymol-psico=3.4.1=py_0
- pyopenssl=19.1.0=py37_0
- pyqt=5.9.2=py37h655552a_2
- pysocks=1.7.1=py37_0
- python=3.7.7=hc70fcce_0_cpython
- python-dateutil=2.8.1=py_0
- pytz=2019.3=py_0
- qt=5.9.7=h468cd18_1
- readline=8.0=h1de35cc_0
- requests=2.23.0=py37_0
- rigimol=1.3=2
- scikit-learn=0.22.1=py37h27c97d8_0
- scipy=1.4.1=py37h9fa6033_0
- setuptools=46.1.3=py37_0
- sip=4.19.8=py37h0a44026_0
- six=1.14.0=py37_0
- sqlite=3.31.1=h5c1f38d_1
- tk=8.6.9=x11tk0_2000
- urllib3=1.25.8=py37_0
- wheel=0.34.2=py37_0
- xz=5.2.5=h1de35cc_0
- zlib=1.2.11=h1de35cc_3
- zstd=1.3.7=h5bba6e5_0
prefix: /Users/jr/anaconda3/envs/rnaposers

Empty file added py/__init__.py
View file
Empty file.
88 changes: 88 additions & 0 deletions py/generate_complex_files.py
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,88 @@
import sys
import inspect
from glob import glob
from pymol import cmd
import psico.fullinit
import tempfile
import os

def generate_complexes(receptor, poses, complexes, ref_complex, ref_ligand_mol2):
def fix_names(obj, is_ligand = False):
# utility functions to fix names
if is_ligand:
cmd.alter("%s"%(obj), "resn = 'UNK'")
cmd.alter("%s"%(obj), "chain = 'Z'")
cmd.alter("%s"%(obj), "resi = '1'")
else:
cmd.alter("%s"%(obj), "type = 'ATOM'")
cmd.alter("resn rC+C+RC and %s"%(obj), "resn = 'CYT'")
cmd.alter("resn rA+A+RA and %s"%(obj), "resn = 'ADE'")
cmd.alter("resn rU+U+RU and %s"%(obj), "resn = 'URA'")
cmd.alter("resn rG+G+RG and %s"%(obj), "resn = 'GUA'")

def split_decoys(poses, dir):
# utility function to split poses
for a in range(1, 1 + cmd.count_states("(%s)"%poses)):
cmd.frame(a)
cmd.save("%s/poses/poses_%i.pdb"%(dir, a))

# utility function to generate complexes
print("[RNAPosers Debugging] Generating complexes...")

# Create TemporaryDirectory to store intermediat files
with tempfile.TemporaryDirectory() as tmpDir:
os.system("mkdir -p %s/complex"%(tmpDir))
os.system("mkdir -p %s/poses"%(tmpDir))

# split posese
cmd.delete("*")
cmd.load(poses, "poses")
fix_names("poses", is_ligand = True)
split_decoys("poses", tmpDir)
poses=range(1, 1 + cmd.count_states("(poses)"))
cmd.delete("poses")

# combine poses + receptor
for a in poses:
cmd.delete("*")
cmd.load(receptor)
cmd.load("%s/poses/poses_%i.pdb"%(tmpDir, a))
cmd.create("test", "all")
fix_names("test", is_ligand = False)
cmd.save("%s/complex/complex_%i.pdb"%(tmpDir, a),"test")

# save dcd
cmd.delete("*")
for a in poses:
cmd.load("%s/complex/complex_%i.pdb" %(tmpDir, a), "complex")
psico.exporting.save_traj(filename = "%s" %complexes, selection = "complex")

# generate reference files
os.system("cp %s/complex/complex_1.pdb %s"%(tmpDir, ref_complex))

# XX
cmd.delete("*")
cmd.load("%s/complex/complex_1.pdb"%(tmpDir), "complex")
cmd.extract("lig", "resn UNK")
cmd.save("%s/complex/lig.pdb"%(tmpDir), "lig")

os.system("obabel -ipdb %s/complex/lig.pdb -omol2 -O %s"%(tmpDir, ref_ligand_mol2))


def main():
# Input: receptor, poses
# Output: complexes, ref_complex, ref_ligand_mol2
# Input
inDir = "test/receptor_and_poses/input/"
receptor = inDir + "receptor.mol2"
poses = inDir + "poses.sd"
# Output
outDir = "test/receptor_and_poses/output/" # where to save your pdb, dcd and mol2 files
os.system("mkdir -p %s" %outDir)
complexes = outDir + "complexes.dcd"
ref_complex = outDir + "complex.pdb"
ref_ligand_mol2 = outDir + "lig.mol2"
generate_complexes(receptor, poses, complexes, ref_complex, ref_ligand_mol2)

if __name__ == '__main__':
main()
30 changes: 13 additions & 17 deletions src/rna_poser.py → py/inference.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -8,27 +8,23 @@
import os
from sklearn.externals import joblib
from sklearn.ensemble import RandomForestClassifier
import argparse

parser = argparse.ArgumentParser()
parser.add_argument("--classifier", help="classifier")
parser.add_argument("--features", help="molecularize features: e.g: /home/itssahil/PROJECTS/Mol_Feturizer/B_LATEST_results/PREDICTORS_R/RF_6bfb.predictor.pkl")
parser.add_argument("--output", help="output scores")

a = parser.parse_args()

def main():
MAXF, MINF = 3.4e38, 1.18e-38
Xtest = numpy.loadtxt(a.features, dtype='float64')
Xtest = Xtest[:, 1:]
# workaround overflow
Xtest = numpy.nan_to_num(Xtest)
Xtest[Xtest < MINF] = MINF
Xtest[Xtest > MAXF] = MAXF
# predict
pred_rf = joblib.load(a.classifier)
rf = pred_rf.predict(Xtest)
rf_p = pred_rf.predict_proba(Xtest)
merge_rf = numpy.column_stack((rf, rf_p))
numpy.savetxt(a.output, merge_rf, fmt='%f')
main()
MAXF, MINF = 3.4e38, 1.18e-38
Xtest = numpy.loadtxt(a.features, dtype='float64')
Xtest = Xtest[:, 1:]
# workaround overflow
Xtest = numpy.nan_to_num(Xtest)
Xtest[Xtest < MINF] = MINF
Xtest[Xtest > MAXF] = MAXF
# predict
pred_rf = joblib.load(a.classifier)
rf = pred_rf.predict(Xtest)
rf_p = pred_rf.predict_proba(Xtest)
merge_rf = numpy.column_stack((rf, rf_p))
numpy.savetxt(a.output, merge_rf, fmt='%f')
Loading

0 comments on commit 866fdd6

Please sign in to comment.