Automating parameter selection to avoid implausible biological pathway models

Abstract

A common way to integrate and analyze large amounts of biological “omic” data is through pathway reconstruction: using condition-specific omic data to create a subnetwork of a generic background network that represents some process or cellular state. A challenge in pathway reconstruction is that adjusting pathway reconstruction algorithms’ parameters produces pathways with drastically different topological properties and biological interpretations. Due to the exploratory nature of pathway reconstruction, there is no ground truth for direct evaluation, so parameter tuning methods typically used in statistics and machine learning are inapplicable. We developed the pathway parameter advising algorithm to tune pathway reconstruction algorithms to minimize biologically implausible predictions. We leverage background knowledge in pathway databases to select pathways whose high-level structure resembles that of manually curated biological pathways. At the core of this method is a graphlet decomposition metric, which measures topological similarity to curated biological pathways. In order to evaluate pathway parameter advising, we compare its performance in avoiding implausible networks and reconstructing pathways from the NetPath database with other parameter selection methods across four pathway reconstruction algorithms. We also demonstrate how pathway parameter advising can guide reconstruction of an influenza host factor network. Pathway parameter advising is method agnostic; it is applicable to any pathway reconstruction algorithm with tunable parameters.

Unveiling Hidden Connections in Omics DataviapyPARAGON: an Integrative Hybrid Approach for Disease Network Construction

“…Here, we only utilized graphlets composed of interactions among 3 and 4 nodes rather than interactions between 2 nodes. However, various graphlet information in reference networks, such as graphlet degree distribution, graphlet frequencies, and probabilistic graphlets, can be embedded in network inference algorithms or biological interpretations 32,34,35,39,40 . However, the use of graphlet features will come with a high computational cost.…”

“…Small connected, nonisomorphic subgraphs, called graphlets, are over-represented in the reference interactome and associated with specific functions 35,36 . Graphlet statistics solve several complex problems in this context, such as the comparison of biological networks, delineating the functional organization of networks, discovering functionally related genes, regulatory interactions, and parameter tuning for network-based approaches 12,32,33,[37][38][39][40] . Another challenge is the presence of highly connected and multifunctional proteins, particularly hub proteins, which can bring nonspecific interactions to the resulting network models because of the small-world property of reference interactomes.…”

Incorporating temporal information during feature engineering bolsters emulation of spatio-temporal emergence

“…Network analysis provides a means to abstract high-resolution, spatial information into summary statistics (Pavlopoulos et al, 2011;Koutrouli et al, 2020). Network topology and morphology have been used to understand ecological systems (Peterson et al, 2013;Modica et al, 2021), interrogate biological pathways (Magnano and Gitter, 2021), analyze neurological structure (Sporns, 2013;Kok et al, 2020;Bassett and Sporns, 2017), and identify novel treatment (Iadevaia et al, 2010). Specifically in tumor development, network analyses are a promising approach to study healthy and pathogenic vascular mimicry and angiogenesis (Amat-Roldan et al, 2015;Alves et al, 2018;Fouladzadeh et al, 2021).…”

Adaptive, sample-specific parameter selection for more accurate transcript assembly

“…Parameter advising has been previously discussed in several bioinformatic areas such as multiple sequence alignment [Cedillo et al, 2022, DeBlasio and Kececioglu, 2014, 2015a,b], biological pathway reconstruction [Magnano and Gitter, 2021] as well as transcript assembly [DeBlasio et al, 2020]. The method proposed in DeBlasio et al [2020] is the first and only work developing a parameter advising system for transcript assembly by learning the advisor set in a data-driven way.…”