Evaluation of reaction gap-filling accuracy by randomization.

Area of Science:

• Systems Biology
• Computational Biology
• Metabolic Engineering

Background:

• Genome-scale metabolic models (GEMs) are crucial for understanding cellular metabolism.
• Computational reaction gap-filling is a common method for completing GEMs.
• The accuracy of automated gap-filling methods is not well-established.

Purpose of the Study:

• To evaluate the accuracy and performance of different reaction gap-filling computational approaches.
• To compare various configurations of the Pathway Tools MetaFlux software for gap-filling.
• To identify optimal parameters for accurate and efficient metabolic model completion.

Main Methods:

• Degraded versions of the EcoCyc-20.0-GEM model were created by removing reactions.
• Gap-filling was performed using Pathway Tools MetaFlux in General Development Mode (GenDev) and Fast Development Mode (FastDev).
• Thirteen GenDev variants were tested, varying linear solvers (SCIP, CPLEX), constraint sets, and MILP methods, assessing accuracy, speed, and information output.

Main Results:

• Significant performance variations were observed across the 13 gap-filling variants.
• The best GenDev variant achieved an average precision of 87% and recall of 61%.
• FastDev showed lower performance with 71% precision and 59% recall; some variants produced invalid or non-minimal solutions.

Conclusions:

• No single gap-filling variant excelled in all aspects; however, one GenDev variant demonstrated a favorable balance of speed, accuracy, and information.
• Despite improvements, approximately 13% of gap-filled reactions were incorrect and 39% were missed, highlighting the continued need for manual curation in metabolic model development.