Pathway design using de novo steps through uncharted biochemical spaces

Abstract

Existing retrosynthesis tools generally traverse production routes from a source to a sink metabolite using known enzymes or de novo steps. Generally, important considerations such as blending known transformations with putative steps, complexity of pathway topology, mass conservation, cofactor balance, thermodynamic feasibility, microbial chassis selection, and cost are largely dealt with in a posteriori fashion. The computational procedure we present here designs bioconversion routes while simultaneously considering any combination of the aforementioned design criteria. First, we track and codify as rules all reaction centers using a prime factorization-based encoding technique (rePrime). Reaction rules and known biotransformations are then simultaneously used by the pathway design algorithm (novoStoic) to trace both metabolites and molecular moieties through balanced bio-conversion strategies. We demonstrate the use of novoStoic in bypassing steps in existing pathways through putative transformations, assembling complex pathways blending both known and putative steps toward pharmaceuticals, and postulating ways to biodegrade xenobiotics.

Expanding Metabolic Capabilities Using Novel Pathway Designs: Computational Tools and Case Studies

“…Depending on the chosen objective, constraint‐based methods predict steady‐state flux solutions that produce the compound of interest with a bias preference toward short pathways with optimal stoichiometry, thermodynamically plausible pathways, or pathways with minimum enzyme cost . Each of these objective choices determines the nature of the optimization program to be solved (linear programming [LP]/nonlinear programming [NLP] or mixed‐integer linear programming [MILP]), and consequently its complexity.…”

“…Assembly of a metabolic reaction network can be achieved by importing metabolic data from curated databases (DBs) and/or from hypothetical reaction DBs derived from generalized reaction rules or molecular signatures . Once the metabolic network has been assembled into a properly charge‐ and mass‐balanced stoichiometric matrix, a diversity of network and pathway prediction methods can be applied for different modeling purposes.…”

“…Notably, these supervised tools rely on generalized reaction rules that can act upon structurally similar metabolites in the DBs, generating new hypothetical reactions, previously unknown, compatible with the Enzyme Commission classification system . More recently, an unsupervised method called rePrime was developed to extract molecular signatures, based on chemical “moieties” from known reactions, and to generate all the feasible, hypothetical reactions enforcing moiety conservation. Importantly, as opposed to previous methods, rePrime also encodes structural information, preventing enzymes to be used interchangeably on stereoisomers that would lead to biologically incorrect predictions.…”

“…Enumeration of all the possible paths (each composed of different enzymatic steps) connecting a set of source metabolites to a target compound is, however, computationally intractable (nondeterministic polynomial‐time [NP]‐hard problem), and so, different requirements and pruning criteria must be employed. For example, criteria such as pathway length, limited number of heterologous or putative enzymes, and thermodynamic realizability have been employed to either reduce the feasible search space or to drive the search to an “optimal” pathway . Notably, the imposition of these criteria for pathway enumeration is not straightforward in all mathematical frameworks (e.g., graph‐based methods) and only stoichiometry‐based optimization methods are capable of naturally accommodating them.…”

“…Notably, the imposition of these criteria for pathway enumeration is not straightforward in all mathematical frameworks (e.g., graph‐based methods) and only stoichiometry‐based optimization methods are capable of naturally accommodating them. Added to recent progress in mathematical frameworks for expanding the biochemical reaction space and the power of state‐of‐the‐art solvers, optimization‐based methods for metabolic pathway prediction emerge as the most promising tools for pathway prospection.…”