inserted function that allows user to add data to graph in the new pandas style (add_data_to_graph_pandas()). This means that it is now possible to add (however many) replicates to the class object. Function was based mostly on read_perturbations_pandas() and add_data_to_graph().
adjusted hysteresis penalty such that now we penalise edges relatively, meaning that if the relative freenrg estimate is high (>1 kcal/mol), the hysteresis penalty does not result in extremely high per-ligand uncertainty estimates.
removed the 0.4 kcal 'bump' (baseline thermal fluctuation) that was applied in all cases, but should actually only be applied when the error estimate is higher than this baseline.
set the balance_hysteresis boolean to be True by default, but it can be set to False when instantiating (i.e. nA = NetworkAnalyser(balance_hysteresis=False)), which will result in stricter per-ligand error estimates.
added a test for this functionality (not sure if I did it right, not very familiar with the infrastructure).
inserted function that allows user to add data to graph in the new pandas style (
add_data_to_graph_pandas()). This means that it is now possible to add (however many) replicates to the class object. Function was based mostly onread_perturbations_pandas()andadd_data_to_graph().adjusted hysteresis penalty such that now we penalise edges relatively, meaning that if the relative freenrg estimate is high (>1 kcal/mol), the hysteresis penalty does not result in extremely high per-ligand uncertainty estimates.
removed the 0.4 kcal 'bump' (baseline thermal fluctuation) that was applied in all cases, but should actually only be applied when the error estimate is higher than this baseline.
set the
balance_hysteresisboolean to be True by default, but it can be set to False when instantiating (i.e.nA = NetworkAnalyser(balance_hysteresis=False)), which will result in stricter per-ligand error estimates.added a test for this functionality (not sure if I did it right, not very familiar with the infrastructure).