Selecting methods for draft GEM generation in multicellular eukaryotes: a comparative analysis

Natalia E Jiménez, Mikael Espinoza, Sebastián Mejías, Sebastián N Mendoza, Ignacia Segovia, J Cristian Salgado, Carlos Conca, Ziomara P Gerdtzen · 2026

Read Selecting methods for draft GEM generation in multicellular eukaryotes: a comparative analysis by Natalia E Jiménez, Mikael Espinoza, Sebastián Mejías, Sebastián N Mendoza, Ignacia Segovia, J Cristian Salgado, Carlos Conca, Ziomara P Gerdtzen (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13390297.txt.

Introduction

Natalia E Jiménez's team compared seven automated tools for building draft genome-scale metabolism-linked models in multicellular eukaryotes, a task most tools were not designed for. They applied AuReMe, CarveMe, merlin, ModelSEED, Pathway Tools, RAVEN, and Reconstructor to a mosquito, a hamster ovary cell line, and a brown alga.

No tool excelled at both model functionality and eukaryotic traits such as compartmentalisation and organism-specific metabolism, with methods trading one strength against the other on network size, consistency, annotation, and speed. The comparison gives teams a practical guide for matching a reconstruction tool to their organism and goal.

Catalog record

CREATORNatalia E Jiménez, Mikael Espinoza, Sebastián Mejías, Sebastián N Mendoza, Ignacia Segovia, J Cristian Salgado, Carlos Conca, Ziomara P Gerdtzen
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE65 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC13390297

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Selecting methods for draft GEM generation in multicellular eukaryotes: a comparative analysis Authors: Natalia E Jiménez, Mikael Espinoza, Sebastián Mejías, Sebastián N Mendoza, Ignacia Segovia, J Cristian Salgado, Carlos Conca, Ziomara P Gerdtzen Published: 2026 ABSTRACT AbstractMotivated by the applications of genome-scale metabolic models (GEMs) for biological discovery and metabolic engineering, several approaches have been developed for automatic generation of draft GEMs. However, most of these methods are not optimized for their use in multicellular eukaryotes and their performance for this task is unclear. In this work we present a comparative analysis of seven automated reconstruction tools (AuReMe, CarveMe, merlin, ModelSEED, Pathway Tools, RAVEN and Reconstructor) applied to three multicellular eukaryotes: the mosquito Aedes aegypti, the CHO (Chinese Hamster Ovary) cell line from Cricetulus griseus and the brown algae Ectocarpus siliculosus. Evaluation of these tools was based on metrics for network size, functionality, consistency, representation of organelle-specific functions and organism-specific metabolites, annotation quality and execution time. We find that methods differ strongly in a trade-off between model functionality and representation of eukaryotic features such as compartmentalization and organism-specific metabolism, with no single approach excelling at both. Our work aims at providing a practical resource to guide researchers in selecting methods for draft generation tailored to organism characteristics and research goals.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12859-026-06455-7.Keywords: Genome-scale models, Automated reconstruction, Multicellular eukaryotes, Metabolic models Background Genome-scale metabolic models (GEMs) are formal representations of the biochemical reactions that can take place in organisms and their requirements on genes and proteins, inferred from annotated genomes [1, 2]. GEMs have emerged as a powerful tool for biological discovery and metabolic engineering, due to their ability to simulate gene knockouts [3–6], describe multi-species relationships [7, 8], contextualize omics data [9, 10], identify gene targets for cancer therapy [11, 12], and provide a framework for mechanistic understanding of phenotypes of interest [13]. The process of manually generating GEMs is complex and time-consuming, often taking up to several years [2]. This fact, alongside the explosion of genomic data driven by modern sequencing technologies [14–17], motivated the development of several automatic strategies for generating draft GEMs from sequenced genomes [18]. These automated tools, rely on specialized biochemical databases (Reconstructor [19], AutoKEGGRef [20], RAVEN 2.0 [21]), their own genome annotation (merlin [22]) or a curated available version including EC numbers to establish a draft version of the desired GEM (Pathway Tools [23], CarveMe [24], ModelSEED, PlantSEED [25, 26], gapseq [27]). Orthology inferences can also be used to retrieve candidate reactions (AuReMe [28]) and information from previously curated models can aid in this process (AuReMe [28], Reconstructor [19], RAVEN2.0 [21]). While these automated approaches have been successfully applied for generating draft GEMs, their primary focus has been on prokaryotes and unicellular eukaryotes such as S. cerevisiae [29] and Y. lipolytica [30]. There are fewer examples of their use in multicellular eukaryotes, with some exceptions such as PlantSEED (based on ModelSEED) and the use of these strategies for generating plant models [22, 25], macroalgae [31] and mammalian cell lines [32]. Automatic drafting of eukaryotic metabolism poses additional challenges related to genome size and complexity [33, 34], intracellular gene location [35, 36], as well as harboring information regarding different cellular types that can display varied metabolic behaviors [37]. Relative performances of different automated reconstruction tools have not been systematically assessed for multicellular eukaryotes. Furthermore, it has not been assessed which model metrics would be relevant to compare for applications of multicellular eukaryotic GEMs. This is a critical issue as metabolic understanding of these organisms has deep implications for public health [38], sustainable economy [31] and the biopharmaceutical industry [32]. In this work, we provide a perspective on the automatic draft GEM reconstruction methods for modeling multicellular eukaryotes. Three organisms with different degrees of prior metabolic characterization were chosen for this comparison: (i) the Chinese Hamster Ovary (CHO) cell line from Cricetulus griseus, which has an extensively curated GEM widely used for basic and applied metabolic research [39–41], (ii) the brown algae Ectocarpus siliculosus whose interactions with its microbial communities have been studied using GEMs [42], and (iii) the mosquito Aedes aegypti, which has a reconstruction used for studying its interactions with its endosymbiont Wolbachia pipientis [38, 43]. In this study, we provide a thorough analysis of the most relevant automated tools for generation of draft GEMs compatible with eukaryotic organisms, highlighting relevant challenges that arise when working with such organisms. Materials and methods Draft generation Three organisms from a wide range of phylogenetic groups and different degrees of knowledge were selected to make this evaluation: (i) Cricetulus griseus and its derived cell line CHO (Chinese Hamster Ovary), for which there is a highly curated genome and GEM [32] and has been widely used in different applications [44, 45], (ii) Ectocarpus siliculosus, a brown kelp for which there is a genome-scale model [31], and (iii) Aedes aegypti, a mosquito for which there is a recent metabolic reconstruction [38]. Cricetulus griseus genome (CriGri_1.0) [46] and its annotation were retrieved from NCBI (GCA_000223135.1). A. aegypti genome (AaegL5.0) [47] and its annotation were retrieved from VectorBase. E. siliculosus genome (ASM31002v1) [48] and its annotation were retrieved from NCBI (GCA_000310025.1). Seven frameworks for generation of draft genome-scale models were retrieved based on literature and their capabilities to handle eukaryotic genomes. The process to generate draft models using these approaches is described below and summarized in Table S1. AuReMe AuReMe (AUtomated REconstruction of MEtabolic models) [28] is a method for reconstruction of metabolic models which allows the incorporation of information included on a template genome-scale model. It starts from an annotated genome of the target processed through Pathway Tools [23] to obtain a draft model. This information is complemented with ortholog data that is mapped to a template model to retrieve additional reactions and gene associations to be included in their draft model. Annotation and orthology-based drafts were generated using AuReMe (version 2.4) as follows: organism-specific Pathway Tools annotation folders (.dat folder) were provided for the annotation-based step, and different reference organisms with their respective models were used for the analyzed organisms in this work. Drosophila melanogaster (Release 6 plus ISO1 MT, GCF_000001215.4) was used as reference organism for A. aegypti [49], Mus musculus (GRCm39, GCF_000001635.27) [50] for CHO (C. griseus) [51] and Chlamydomonas reinhardtii (Chlamydomonas_reinhardtii_v5.5, GCF_000002595.2) [52] for E. siliculosus [53]. Additional reconstructions were obtained using Homo sapiens (GRCh38.p14, GCF_000001405.40) [54] for A. aegypti and CHO, and Saccharina japonica [28, 55] for E. siliculosus to test the effect of reference organisms in the obtained quality of reconstructions. CarveMe CarveMe [24] starts from an universal model which is carved based on information contained on its proteome. This method, originally developed from prokaryotes, proposed a biomass adapted from E. coli, and removes blocked reactions and dead-end metabolites to retrieve a functional model. In this work, CarveMe [24] was used with amino acid fasta files of the target organisms as input using their default settings (bacterial template and the arguments—more-sensitive—top 10 for DIAMOND [56]). Merlin Merlin [22] is a java application with its own user interface that performs its own annotation, transport and compartment location predictions and retrieves a draft model that can be curated by the user. It performs enzyme annotation on its own server

Read the complete text