You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A logistic regression model was used for binary classification to predict edges between a given pair of nodes.
139
139
The logistic regression is one of the most popular statistical models used for binary classification because it is simple, easy to interpret, and proven to perform well in many classification and prediction tasks.
140
-
The results of the classifier are interpreted using two measures, $p$-value and minus log probability.
141
-
The $p$-value is the probability that the null hypothesis is true.
142
-
In this case, it is that a given edge does not exist between two nodes in the graph.
143
-
The \ac{MLP} is also reported to enable easier interpretation.
144
-
Generally, a $p$-value of less than 0.05, corresponding to an \ac{MLP} of greater than 1.3, is considered to be significant.
145
-
It is worth mentioning that in this thesis, the $p$-value is mainly for ranking purposes, and multiple hypothesis testing correction was not employed.
140
+
The results of the classifier were ranked using log odds ratio (LOR). The lower LOR is, the higher is the probability of the prediction to be true.
Copy file name to clipboardExpand all lines: content/results.tex
+49-49Lines changed: 49 additions & 49 deletions
Original file line number
Diff line number
Diff line change
@@ -122,7 +122,7 @@ \section{Interpretation of Model Predictions}
122
122
The predictions were created by querying the name or identifier of a certain entity.
123
123
Furthermore, the type of entities to be predicted could also be specified.
124
124
Three types of relations could be predicted using the model: drug-phenotype, drug-target, and target-phenotype.
125
-
For each relation predicted, the $p$-value and \ac{MLP} value were calculated and used for ranking.
125
+
For each relation predicted, the LOR was calculated and used for ranking.
126
126
Since methods for drug-target associations, or drug target identification, are well developed, those association results were omitted from the interpretations.
127
127
To validate the new predictions of the model, positive controls were also presented for drug-phenotype associations.
128
128
@@ -135,29 +135,29 @@ \subsection{Predicting the Phenotypes for a Drug}
0 commit comments