schrodinger.application.desmond.correlation_tau module¶

This program computes the average correlation coefficient and the kendall tau rank coefficient between experiment and prediction samples. Samples are randomly drawn from gaussian distribution centered on each experimental data point with given error.

The algorithm used in this program is based on the work by Scott P. Brown, Steven W. Muchmore, Philip J. Hajduk Drug Discovery Today, Vol. 14, No. 7-8., pp. 420-427

class schrodinger.application.desmond.correlation_tau.ConfidenceInterval(val: float, lower_bound: float, upper_bound: float)¶

Bases: object

val: float¶

lower_bound: float¶

upper_bound: float¶

to_measurement() → schrodinger.application.desmond.measurement.Measurement¶

Convert the ConfidenceInterval into a Measurement

Since confidence intervals (ci) are non-symetrical, we take its average to make it symmetrical. We then divide it by two to make it same scale as standard deviation, assuming that the input is a 95% confidence interval.

Returns: measurement derived from the represented confidence interval

__init__(val: float, lower_bound: float, upper_bound: float) → None¶

schrodinger.application.desmond.correlation_tau.predict_kendall_tau(experiment, experiment_sigma=0.3, prediction_sigma=0.3, num_sample=1000)¶

Computes the average Kendall tau rank correlation coefficient between experiment and prediction samples. num_sample independent data for each experiment and prediction are sampled from gaussian distribution with experiment_sigma and prediction_sigma error.

Parameters

experiment – sequence of experiment data
experiment_sigma – experimental error
prediction_sigma – prediction error
num_sample – number of samples

Returns

average_tau, sigma_tau

schrodinger.application.desmond.correlation_tau.predict_expected_slope(experiment, experiment_sigma=0.3, prediction_sigma=0.3, num_sample=1000)¶

schrodinger.application.desmond.correlation_tau.predict_correlation(experiment, experiment_sigma=0.3, prediction_sigma=0.3, num_sample=1000, return_ci=False)¶

Computes the average correlation coefficient between experiment and prediction samples. num_sample independent data for each experiment and prediction are sampled from gaussian distribution with experiment_sigma and prediction_sigma error.

Parameters

experiment – sequence of experiment data
experiment_sigma – experimental error
prediction_sigma – prediction error
num_sample – number of samples
return_ci – If True, return the data in ConfidenceInterval format: <R>, <R^2>, <R^2_signed> If False, return: <R>, sigma_R, <R^2>, sigma_R^2, <R^2_signed>, sigma_R^2_signed

schrodinger.application.desmond.correlation_tau.compute_rmse(experiment, prediction)¶

Computes root mean square error between experiment and prediction. Averages of experiment and prediction are aligned before RMSE computation.

Parameters

experiment – sequence of experiment data
prediction – sequence of prediction data

Returns

root_mean_square_error between experiment and prediction