def transfer_energy(atoms, sas):
"""Calculate transfer energy for each osmolyte and amino acid type.
.. todo:: implement max/min for Creamer model
:param list(pqr.Atom) atoms: list of atoms
:param SolventAccessibleSurface sas: solvent-accessible surface area object
:returns: DataFrame with energy contribution (kcal/mol) by residue type
:rtype: pd.DataFrame
"""
ref_model = ReferenceModels()
counts = count_residues(atoms)
folded_areas = get_folded_areas(atoms, sas)
unfolded_areas = ref_model.molecule_areas(atoms, model="auton", how="mean")
tripeptide_areas = ref_model.molecule_areas(atoms,
model="tripeptide",
how="mean")
scaled_areas = (unfolded_areas - folded_areas) / tripeptide_areas
scaled_areas["sidechain"]["GLY"] = 0.0
osmolytes = list(ENERGY_DICT["backbone"].keys())
residue_energies = {}
bb_energy = {osmolyte: 0.0 for osmolyte in osmolytes}
for res, area in scaled_areas.T.iteritems():
sc_energy = {}
for osmolyte in osmolytes:
sc_energy[osmolyte] = (counts[res] * ENERGY_DICT[res][osmolyte] *
area["sidechain"])
bb_energy[osmolyte] += (counts[res] *
ENERGY_DICT["backbone"][osmolyte] *
area["backbone"])
residue_energies[res] = sc_energy
residue_energies["backbone"] = bb_energy
return pd.DataFrame(residue_energies).T.sort_index() / 1000
def main(args=None):
"""Main entry point.
:param list(str) args: list of strings to parse
:returns: dictionary of results
:rtype: dict
"""
parser = build_parser()
args = parser.parse_args(args)
logging.basicConfig(level=getattr(logging, args.log_level))
_LOGGER.info(f"Osmolytes version {osmolytes.__version__}")
_LOGGER.debug(f"Got command-line arguments: {args}")
_LOGGER.info(f"Reading PQR input from {args.pqr_path}...")
with open(args.pqr_path, "rt") as pqr_file:
atoms = parse_pqr_file(pqr_file)
_LOGGER.info(
f"Constructing protein solvent-accessible surface with "
f"solvent radius {args.solvent_radius} and {args.surface_points} "
f"points per atom."
)
sas = SolventAccessibleSurface(
atoms,
probe_radius=args.solvent_radius,
num_points=args.surface_points,
xyz_path=args.surface_output,
)
counts = count_residues(atoms)
areas = get_folded_areas(atoms, sas).rename(
{"sidechain": "sidechain areas", "backbone": "backbone areas"},
axis="columns"
)
areas["residue counts"] = counts
print(areas)
areas = areas[["residue counts", "sidechain areas", "backbone areas"]]
_LOGGER.info(f"Protein composition:\n{areas}")
_LOGGER.info("Calculating transfer energies.")
energy_df = transfer_energy(atoms, sas)
_LOGGER.info(f"Detailed energies (kcal/mol/M):\n{energy_df.to_string()}")
energies = energy_df.sum(axis=0).sort_values()
_LOGGER.info(f"Summary energies (kcal/mol/M):\n{energies}")
output_dir = Path(args.output_dir)
if args.output == "xlsx":
output_path = output_dir / f"{Path(args.pqr_path).stem}-mvalues.xlsx"
_LOGGER.info(f"Writing energies (kcal/mol/M) to {output_path}")
energy_df.to_excel(output_path)
elif args.output == "csv":
output_path = output_dir / f"{Path(args.pqr_path).stem}-mvalues.csv"
_LOGGER.info(f"Writing energies (kcal/mol/M) to {output_path}")
energy_df.to_csv(output_path)
return {
"args": args,
"atoms": atoms,
"sas": sas,
"energy_df": energy_df,
"energies": energies,
}
def test_residue_count(protein):
"""Test aggregate SASAs per residue type as reported in Auton and Bolen
(doi:10.1073/pnas.0507053102, Supporting Table 2)."""
pqr_path = PROTEIN_PATH / f"{protein}.pqr"
with open(pqr_path, "rt") as pqr_file:
atoms = parse_pqr_file(pqr_file)
test_counts = count_residues(atoms).sort_index()
ref_path = PROTEIN_PATH / "Auton-Bolen.yaml"
with open(ref_path, "rt") as ref_file:
ref_dict = yaml.load(ref_file, Loader=yaml.FullLoader)
ref_dict = ref_dict[protein]
ref_df = pd.DataFrame(ref_dict).T
ref_df = ref_df[ref_df["number"] > 0]
ref_counts = np.array(ref_df["number"].sort_index())
results_df = pd.DataFrame({"Test": test_counts, "Ref": ref_counts})
_LOGGER.info(f"Results for {protein}:\n{results_df}")
np.testing.assert_equal(test_counts, ref_counts)
def test_auton_sasa(protein, tmp_path):
"""Test aggregate SASAs per residue type as reported in Auton and Bolen
(doi:10.1073/pnas.0507053102, Supporting Table 2)."""
_LOGGER.info("Temp path: %s", tmp_path)
pqr_path = PROTEIN_PATH / f"{protein}.pqr"
with open(pqr_path, "rt") as pqr_file:
atoms = parse_pqr_file(pqr_file)
xyz_path = Path(tmp_path) / "surface.xyz"
sas = SolventAccessibleSurface(
atoms, probe_radius=1.4, num_points=200, xyz_path=xyz_path
)
test_areas = [sas.atom_surface_area(iatom) for iatom in range(len(atoms))]
test_areas = aggregate(
atoms,
test_areas,
chain_id=False,
res_name=True,
res_num=False,
sidechain=True,
)
test_areas = test_areas.set_index("res_name").sort_index()
test_counts = np.array(count_residues(atoms))
ref_path = PROTEIN_PATH / "Auton-Bolen.yaml"
with open(ref_path, "rt") as ref_file:
ref_dict = yaml.load(ref_file, Loader=yaml.FullLoader)
ref_dict = ref_dict[protein]
ref_df = pd.DataFrame(ref_dict).T
ref_df = ref_df[ref_df["number"] > 0]
ref_counts = np.array(ref_df["number"].sort_index())
np.testing.assert_equal(test_counts, ref_counts)
for group in ["backbone", "sidechain"]:
ref = np.array(ref_df[group])
test = np.array(test_areas[test_areas["sidechain"] == group]["value"])
# Test correlation since we're not 100% sure what parameters were used
# in PNAS paper
results = linregress(ref, test)
info = (
f"{group} areas are correlated with results from paper: "
f"{results}"
)
_LOGGER.info(info)
if not np.isclose(results.pvalue, 0):
err = f"Poor fit to {group} areas: {results}"
raise AssertionError(err)
def molecule_areas(self, atoms, model, how="mean"):
"""Return reference areas for the molecule.
:param list(pqr.Atom) residue: list of atoms in molecule
:param str model: model to use
- creamer: denatured areas from Creamer, Srinivasan, Rose.
doi: 10.1021/bi962819o
- auton: denatured areas from Auton, Bolen
doi: 10.1073/pnas.0507053102
- tripeptide: Gly-X-Gly areas for residue X from Auton, Bolen
doi: 10.1073/pnas.0507053102
:param str what: what area to report
- sidechain
- backbone
- total
:param str how: calculation to perform on model values
- max
- min
- mean
:returns: DataFrame with a row for each residue type
:rtype: pd.DataFrame
"""
count = count_residues(atoms)
rows = {}
for res, num in count.iteritems():
res_areas = self.residue_area(res, model, how)
row = {}
for what in ["sidechain", "backbone"]:
row[what] = num * res_areas[what]
rows[res] = row
return pd.DataFrame(rows).T.sort_index()