• Overview
• Affinity Isolation
• Affinity Purification
• Data Mining »
  • Vizualisations
  • Bioinformatics
  • Bead Proteomes & the PFL

For each SILAC IP experiment, the distribution of ratios can be plotted as a histogram. Thus, for either M/L, H/L; or H/M SILAC ratios, the number of proteins with each ratio value is plotted on the y axis against log₂ SILAC ratio values on the x axis. Non-specific, experimental contaminants, which can represent more than 80% of the total number of proteins identified), reproducibly cluster in a Gaussian (normal) distribution centered at the log2 ratio of ~0 (which corresponds to a SILAC ratio of ~1). Theoritically, the normal distribution should be centered on a log2 value of exactly 0, but, in practice, this varies between individual experiments and the actual mean can be either higher or lower even for the separate M/L, H/L and H/M ratios measured within a single triple SILAC experiment. This graph shows that a minor group of proteins has a high SILAC M/L ratio (log2ratio > 1, specific protein interaction partners), whereas 80% of the proteins have a log2ratio between -1.4 and 1.4 that can be considered as putative experimental contaminants.

Another way of visualizing the data of a triple SILAC co-IP experiment is to plot log₂(H/M) (y axis) versus log₂(M/L) (x axis) SILAC ratio values for all proteins identified. This visualization provides an indication of both the specificity of the interaction (M/L ratio) and the changes occurring between the two conditions tested (H/M ratio). In this graph, putative experimental contaminants are clustered around the origin, with both log₂(M/L) and log₂(H/M) of approximately 0. Because of experimental variability, the cluster of contaminants can however deviate from the origin. In this case, it is recommended that the IP dataset is "normalized", with the median SILAC ratio value of the predicted group of contaminants being set to exactly zero (see details below and in (Boulon, Ahmad et al. 2010)). In contrast, the bait protein and its putative specific interaction partners are present on the right side of the graph. If the pull-down efficiency is the same between the two conditions tested, the log₂(H/M) ratio should be 0 for the bait protein. In practice, this is often not the case due for example to variations in expression levels, accessibility and/or fractionation efficiency induced by the treatment. Hence, the bait protein can be used as a reference point to draw a second x axis such that proteins falling above the new x axis line indicate increased interaction with the bait and proteins falling below indicate decreased interaction as a result of the treatment.

Quantitative label-free methods can also be applied to the reliable identification of specific interaction partners, as long as the quantitation is robust and reproducible (Hubner, Bird et al. 2010).

However, regardless of how the MS pull-down data are visualized, a significant overlap invariably exists between the ratio values of specific interaction partners and contaminating proteins and it is often impossible to establish a ratio threshold for pull-down experiments that eliminates all of the non specific background without discarding, en passant, genuine interaction partners of lower abundance and/or lower binding affinity. This means that relying upon isotope labelling ratios alone (or quantitative label-free methods) does not entirely solve the contaminant problem. To address this issue, objective criterion need to be added to the analysis.