Learning signaling networks from combinatorial perturbations by exploiting siRNA off-target effects

Area of Science:

• Computational biology
• Systems biology
• Network inference

Background:

• Perturbation experiments are crucial for studying cellular networks but are complicated by factors like RNA interference (RNAi) off-target effects.
• These effects lead to combinatorial gene perturbations and signal propagation through unknown network connections, complicating computational modeling.
• Understanding how many and which perturbations are needed to accurately infer signaling network models is a key challenge.

Purpose of the Study:

• To introduce an enhanced version of linear effects models (LEMs) capable of accounting for both positive and negative contributions of perturbed proteins.
• To determine the minimal perturbation combinations required for unique and accurate network model identification.
• To improve the accuracy of parameter estimation and network structure learning from perturbation data.

Main Methods:

• Developed an enhanced linear effects model (LEM) incorporating positive and negative protein contributions.
• Proved identifiability of enhanced LEMs using data from single, pair, and triplet protein perturbations.
• Validated the model's performance through extensive simulations and application to real-world RNAi screening data.

Main Results:

• Enhanced LEMs are identifiable with perturbations of single, pairs, and triplets; only single and pairs are needed for small networks (≤5 nodes).
• Extensive simulations show enhanced LEMs achieve high accuracy in parameter estimation and network structure learning, outperforming previous models.
• Application to Bartonella henselae infection data identified known interactions and suggested a novel interaction.

Conclusions:

• Enhanced LEMs provide a robust framework for inferring signaling networks from complex perturbation data.
• The model accurately estimates parameters and learns network structures, even with RNAi off-target effects.
• This approach advances the computational modeling of cellular signaling pathways and aids in discovering biological interactions.