Direct and indirect interactions in AP-MS data sets. The diagram depicts a bait protein bound to a prey protein complex. Solid lines indicate bait-prey interactions that could be observed in an AP-MS experiment, while dashed lines indicate prey-prey protein complex interactions that are not directly observable. Spoke methods make predictions only on directly observed interactions (e.g., Bait with Prey A), whereas matrix models infer protein complex interactions (e.g., Prey A with Prey B). Because the prey proteins do not necessarily form a single complex that interacts with the bait, inferences of prey-prey interactions need to be based on the co-occurrence of pairs of preys across many purification experiments, which strengthens the evidence for interaction.