A critical comparison of topology-based pathway analysis methods.

摘要: One of the aims high-throughput gene/protein profiling experiments is identification biological processes altered between two or more conditions. Pathway analysis an umbrella term for a multitude computational approaches used this purpose. While in beginning pathway relied on enrichment-based approaches, newer generation methods now available, exploiting topologies addition to expression levels. However, little effort has been invested their critical assessment with respect performance different experimental setups. Here, we assessed seven representative identifying differentially expressed pathways groups interest based gene data prior knowledge topologies: SPIA, PRS, CePa, TAPPA, TopologyGSA, Clipper and DEGraph. We performed number controlled that investigated sensitivity sample size, threshold-based filtering genes, ability detect target pathways, exploit topological information pre-processing strategies. also verified type I error rates described influence overexpression single sets motifs various sizes detection as expressed. The results our demonstrate wide variability tested methods. provide set recommendations informed selection proper method given task.