Source code for schrodinger.test.stu.outcomes.correlate_to_reference_file

import numpy as np
import scipy.stats.stats

from schrodinger import structure


def read_prop_by_title(fn, prop_name):
    """
    Read data from a structure file into a dictionary by title
    fn {string} file to read
    prop_name {string} ct-level property to read
    """
    data = {}
    # Get the data from file_name
    for ct in structure.StructureReader(fn):
        if prop_name in ct.property:
            if ct.title in data:
                raise RuntimeError('Titles must be unique in %s (%s)' %
                                   (fn, ct.title))
            data[ct.title] = ct.property[prop_name]
    return data


def correlate_to_reference_file_workup(file_name, file_property, ref_file_name,
                                       ref_file_property, min_correl):
    r"""
    Determine the correlation between ct-level real properties from one
    file to a reference.  CT's are matched by their titles.

    :param file_name: Path to the file to check, readable by structurereader
    :type file_name: str

    :param file_property: Name of the ct-level property in the file_name to
                          use. This should start with r\_ or i\_
    :type file_property: str

    :param ref_file_name: Path to the reference file, readable by
                          structurereader
    :type ref_file_name: str

    :param ref_file_property: Name of the ct-level property in ref_file_name to
                              use. This should start with a r\_ or i\_
    :type ref_file_property: str

    :param min_correl: Correlation (R, not R-squared) should be less than this
                       value if this is a negative value (more negative
                       correlation) and more than this value if this is a
                       positive value (more positive correlation)
    :type min_correl: float
    """
    # Match by title
    file_dict = read_prop_by_title(file_name, file_property)
    ref_dict = read_prop_by_title(ref_file_name, ref_file_property)
    # Create lists of the data
    ref_data = []
    file_data = []
    for title in ref_dict:
        if (title not in file_dict):
            raise RuntimeError("Cannot find data for %s(%s) in %s" %
                               (title, file_property, file_name))
        ref_data.append(ref_dict[title])
        file_data.append(file_dict[title])
    # Get the correlation
    results = scipy.stats.stats.linregress(np.array(file_data),
                                           np.array(ref_data))
    correl = results[2]
    report = f"Correlation is {correl:5.3f} -- Target is {min_correl:5.3f}"
    if min_correl < 0 and correl > min_correl:
        raise AssertionError(report)
    elif min_correl > 0 and correl < min_correl:
        raise AssertionError(report)
    return True