Comparison of the performance between <i>P</i>, <i>P</i>_<i>S</i> and <i>P</i>_<i>SI</i>.

<p>In both plots, <i>x</i>-axes are the number (<i>r</i>) of top-ranked protein pairs (ranked by their probabilities: <i>P</i>, <i>P</i>_<i>S</i> and <i>P</i>_<i>SI</i>); <i>y</i>-axes are the KEGG (<i>a</i>) and the GO (<i>b</i>) annotation overlap rates–<b><i>Q</i></b>_{<b><i>k</i></b>}(<b><i>r</i></b>) and <b><i>Q</i></b>_{<b><i>g</i></b>}(<b><i>r</i></b>) for the top-ranked <i>r</i> protein pairs. Line colors represent the three formulas: green for (1), blue for (2) and red for (3). Dotted black lines (between green and blue lines) represent formula (2) with direct links integrated (with <i>λ</i> = 2). Vertical dashed lines (<i>r</i> = 8,583) represent the cut-off for significantly associated protein pairs. Fig.3 is based on the top-ranked 30,000 protein pairs from the three lists (each consists of over 1.5 million protein pairs).