Ligify: Automated genome mining for ligand-inducible transcription factors

Simon d’Oelsnitz, Andrew D. Ellington, David Ross

TL;DR

Researchers have developed Ligify, a web tool that automates the discovery and design of biosensors for detecting specific chemicals. By leveraging enzyme reaction databases, Ligify predicts transcription factors that respond to user-defined molecules and automatically generates the genetic circuits needed to build these biosensors, significantly accelerating the process of creating new chemical sensors.

What Problem Does This Paper Address?

Developing biosensors, particularly those that respond to specific chemical signals, is crucial for various applications in synthetic biology, biotechnology, and diagnostics. However, the number of known chemical ligands for which functional biosensors exist is very limited. The traditional process of identifying and engineering these biosensors is laborious, involving manual searching of literature and databases for transcription factors that respond to a target chemical. This bottleneck hinders the rapid expansion of our toolkit for chemical sensing. This paper addresses this limitation by presenting Ligify, a tool designed to automate and streamline the discovery and design of these valuable biosensors.

Methods

Ligify employs a "guilt-by-association" approach. The core workflow involves:

  1. Chemical Input: Users provide a small molecule of interest (as SMILES, text, or drawing).
  2. Reaction Association: Ligify queries the Rhea database to find enzymatic reactions involving this molecule.
  3. Enzyme Identification: It then identifies bacterial enzymes from UniProt associated with these reactions, filtering for unique and relevant enzymes.
  4. Operon Mining: For each enzyme, Ligify retrieves its genomic context (operon) from NCBI. It searches within these operons for genes annotated as "regulators," "repressors," or "activators."
  5. Promoter Prediction: The region between divergently expressed gene sets within the operon is hypothesized to contain a promoter regulated by the identified transcription factor.
  6. Ranking: A heuristic scoring system ranks potential regulators based on factors like proximity to the associated enzyme and operon characteristics.
  7. Plasmid Design: For promising candidates, Ligify automatically designs a plasmid containing the regulator and a reporter gene (GFP), driven by a constitutive promoter for the regulator and the predicted inter-operon promoter for GFP.

Datasets

  • Rhea Database: Used to identify enzymatic reactions associated with input chemicals.
  • UniProt: Used to find bacterial enzymes involved in those reactions.
  • NCBI: Used to retrieve bacterial operon sequences.
  • groovDB: A curated dataset of 100 experimentally validated ligand-regulator pairs used for benchmarking Ligify and comparing it to TFBMiner. This dataset represents existing knowledge of ligand-inducible transcription factors.

Key Insights and Ideas

  • Automated Biosensor Design is Feasible: Ligify demonstrates that a significant portion of the biosensor discovery and design pipeline can be automated, drastically reducing the time and effort required.
  • "Guilt-by-Association" is Powerful but Imperfect: The core strategy of linking chemical response to co-expressed enzymes is effective but relies heavily on the completeness of reaction and genomic databases.
  • Database Choice Matters: The comparison with TFBMiner highlights that the choice of reaction database (Rhea vs. KEGG) can significantly impact prediction accuracy, suggesting that integrating multiple databases could further improve performance.
  • Regulator-Enzyme Proximity and Operon Structure are Good Heuristics: The ranking system's reliance on these factors suggests that functionally related genes are often clustered in bacterial genomes, providing a useful signal for identifying regulatory relationships.
  • Ligify Generalizes Across Regulator Families: The successful creation of biosensors from regulators belonging to distinct structural families indicates the tool's broad applicability.

Key Results

  • Benchmarking Accuracy: Ligify correctly predicted 31 out of 100 previously validated biosensor:ligand pairs, outperforming TFBMiner (26/100).
  • Novel Biosensor Creation: Ligify was used to design and experimentally validate four novel biosensors for D-Ribose, L-Sorbose, Isoeugenol, and 4-Vinylphenol.
  • Induction Ratios: These novel biosensors achieved significant induction ratios, including 47-fold for D-Ribose, 5-fold for L-Sorbose, 9-fold for Isoeugenol, and 27-fold for 4-Vinylphenol.
  • Dose-Response Characterization: The tool facilitated the characterization of dose-dependent responses and selectivity for several regulators.

Limitations and Future Work

  • Database Completeness: A major limitation is the reliance on the completeness of reaction and genomic databases. Missing information in these databases leads to prediction failures.
  • Search Algorithm: The current search algorithm for regulators within operons could be improved to capture more candidates, especially those with complex genetic contexts.
  • Promoter and RBS Tuning: The generic plasmid design used for all regulators, without specific promoter or Ribosome Binding Site (RBS) tuning, might limit the performance of some biosensors.
  • Future Work:
    • Database Integration: Incorporating additional reaction databases (e.g., BRENDA, Expasy) to improve coverage.
  • Algorithm Refinement: Enhancing the operon search algorithm to better identify regulators.
  • Advanced Ranking: Developing more sophisticated ranking systems that consider factors like transmembrane domains or enzyme promiscuity.
  • "Reverse" Ligify: Developing a mode to suggest potential effector ligands for a given regulator.
  • Machine Learning Integration: Using Ligify-generated data to train machine learning models for predicting regulator:ligand pairings.
  • Operator Prediction: Integrating with tools like Snowprint to predict DNA binding sites (operators) for identified regulators.

Summary updated

Abstract

Abstract Abstract Figure Prokaryotic transcription factors can be repurposed into biosensors for the ligand-inducible control of gene expression, but the landscape of chemical ligands for which biosensors exist is extremely limited. To expand this landscape, we developed Ligify, a web application that leverages information in enzyme reaction databases to predict transcription factors that may be responsive to user-defined chemicals. Candidate transcription factors are then incorporated into automatically generated plasmid sequences that are designed to express GFP in response to the target chemical. Our benchmarking analyses demonstrated that Ligify correctly predicted 31/100 previously validated biosensors, and highlighted strategies for further improvement. We then used Ligify to build a panel of genetic circuits that could induce a 47-fold, 5-fold, 9-fold, and 27-fold change in fluorescence in response to D-ribose, L-sorbose, isoeugenol, and 4-vinylphenol, respectively. Ligify should enhance the ability of researchers to quickly develop biosensors for an expanded range of chemicals, and is publicly available at https://ligify.streamlit.app .
BibTeX
@inproceedings{doelsnitz2024ligify,
  author = {Simon d’Oelsnitz and Andrew D. Ellington and David Ross},
  title = {{Ligify: Automated genome mining for ligand-inducible transcription factors}},
  booktitle = {ACS},
  year = {2024},
  doi = {10.1021/acssynbio.4c00372},
}
Figure