BACKGROUND: Sequence comparison is one of the most prominent tools in biological research, and is instrumental in studying gene function and evolution. The rapid development of high-throughput technologies for measuring protein interactions calls for extending this fundamental operation to the level of pathways in protein networks. RESULTS: We present a comprehensive framework for protein network searches using pathway queries. Given a linear query pathway and a network of interest, our algorithm, QPath, efficiently searches the network for homologous pathways, allowing both insertions and deletions of proteins in the identified pathways. Matched pathways are automatically scored according to their variation from the query pathway in terms of the protein insertions and deletions they employ, the sequence similarity of their constituent proteins to the query proteins, and the reliability of their constituent interactions. We applied QPath to systematically infer protein pathways in fly using an extensive collection of 271 putative pathways from yeast. QPath identified 69 conserved pathways whose members were both functionally enriched and coherently expressed. The resulting pathways tended to preserve the function of the original query pathways, allowing us to derive a first annotated map of conserved protein pathways in fly. CONCLUSION: Pathway homology searches using QPath provide a powerful approach for identifying biologically significant pathways and inferring their function. The growing amounts of protein interactions in public databases underscore the importance of our network querying framework for mining protein network data.

No abstract available.

This paper is a continuation of our previous work in [12]. The results, and some applications, have been described in the announcement [13]; it may be useful to discuss here, a little more fully, the nature and purpose of this work. We are concerned basically with three kinds of algorithmic problem: (1) isomorphism problems, (2) “orbit problems”, and (3) “effective generation”. (1) Isomorphism problems. Here we have a class of algebraic objects of some kind, and ask: is there a uniform algorithm for deciding whether two arbitrary members of are isomorphic? In most cases, the answer is no: no such algorithm exists. Indeed this has been one of the most notable applications of methods of mathematical logic in algebra (see [26, Chapter IV, §4] for the case where is the class of all finitely presented groups). It turns out, however, that when consists of objects which are in a certain sense “finite-dimensional”, then the isomorphism problem is indeed algorithmically soluble. We gave such algorithms in [12] for the following cases: = {finitely generated nilpotent groups}; = {(not necessarily associative) rings whose additive group is finitely generated}; = {finitely Z-generated modules over a fixed finitely generated ring}. Combining the methods of [12] with his own earlier work, Sarkisian has obtained analogous results with the integers replaced by the rationals: in [20] and [21] he solves the isomorphism problem for radicable torsion-free nilpotent groups of finite rank and for finite-dimensional Q-algebras.