this is for holding javascript data
Timothy O'Donnell edited section_Evaluating_the_performance_of__.tex
almost 8 years ago
Commit id: 2b15eecbcd53336b2b8feb1e2bfadfd4c6764f04
deletions | additions
diff --git a/section_Evaluating_the_performance_of__.tex b/section_Evaluating_the_performance_of__.tex
index 4253f88..b75ab97 100644
--- a/section_Evaluating_the_performance_of__.tex
+++ b/section_Evaluating_the_performance_of__.tex
...
\section{Evaluating the performance of a binding predictor}
Two datasets were used from a recent paper studying the relationship between training data and pMHC predictor accuracy\cite{Kim_2014}. The training dataset (BD2009) contained entries from IEDB\cite{Salimi_2012} up to 2009 and the test dataset (BLIND) contained IEDB entries from between 2010 and 2013 which did not overlap with BD2009 (Table~\ref{tab:datasets}).
\section{Evaluating the performance of a binding predictor} \begin{table}[h!]
\centering
\begin{tabular}{l||cccc}
\toprule
{} & Alleles & IC50 Measurements & Expanded 9mers \\
\midrule
BD2009 & 106 & 137,654 & 470,170 \\
BLIND & 53 & 27,680 & 83,752 \\
\bottomrule
\end{tabular}
\caption{Train (BD2009) and test (BLIND) dataset sizes.}
\label{tab:datasets}
\end{table}
Throughout this paper we will evaluate a pMHC binding predictor using three different metrics:
\begin{itemize}
\item {\bf F$_1$ score}: Measures trade-off between sensitivity and specificity for predicting ``strong binders'' with affinities $<= 500$nM.
\item {\bf AUC}: Area under the ROC curve. Estimates the probability that a ``strong binder'' peptide will be given a stronger predicted affinity than one whose ground truth affinity is $>500$nM.
\item {\bf Kendall's $\tau$}: Rank correlation across the full spectrum of binding affinities.