arXiv subjects

Jean-Loup Faulon

Publications and source records attributed to Jean-Loup Faulon.

5 recordsLinked to original sources

arXiv2026

An Agentic Retrobiosynthesis Framework with Learned Frontier Selection

Large language models are increasingly used as agents for multistep retrosynthesis, raising the question of how much their search policy contributes independently of the underlying reaction model. We investigate this question in a biological setting through rule-based retrobiosynthesis: a deterministic biochemical engine generates the same validated transitions for every method, searching for routes that terminate in metabolites available to an \emph{Escherichia coli} chassis, while the policy only selects which frontier molecule to expand next. Prompted and LoRA-tuned Qwen2.5-7B policies use a strict choice-only interface. The fine-tuned policy reaches $65\pm1$\% solve rate at 10 expansions on LASER versus 59\% for MCTS, and at 200 expansions reaches $78\pm1$\% versus 75\% on LASER, $88\pm3$\% versus 80\% on the RetroPath RL Golden benchmark, and $63\pm2$\% versus 45\% on the BioNavi-NP benchmark. Fine-tuning also consistently outperforms direct prompting. These results show that route-supervised frontier selection can improve budgeted search without altering biochemical generation, although performance remains dependent on frontier construction and reaction ranking.

cs.CL↗

arXiv2026

What Makes a Bacterial Model a Good Reservoir Computer? Predicting Performance from Separability and Similarity

Biological systems are promising substrates for computation because they naturally process environmental information through complex internal dynamics. In this study, we investigate whether bacterial metabolic models can act as physical reservoirs and whether their computational performance can be predicted from dynamical properties linked to separability and similarity. We simulated the growth dynamics of five bacterial species, one yeast species, and 29 Escherichia coli single-gene deletion mutants using dynamic flux balance analysis (dFBA), with glucose and xylose concentrations as inputs and growth curves as reservoir states. Computational performance was assessed on random nonlinear classification tasks using a linear readout, while reservoir properties linked to separability and similarity were characterised through kernel and generalisation ranks computed from growth-curve state matrices. Several microbial models achieved high classification accuracy, showing that bacterial metabolic dynamics can support nonlinear computation. Clear differences were observed between species, with some models converging more rapidly and others reaching higher maximum accuracy, revealing a trade-off between convergence speed and peak performance. In contrast, all E. coli mutants were dominated by the wild-type model, suggesting that gene deletions reduce the dynamical richness required for efficient computation. The difference between kernel and generalisation ranks was generally associated with improved accuracy, but deviations across models and sensitivity at low rank values limited its predictive power in practice. Overall, these results show that bacterial metabolic models constitute promising substrates for reservoir computing and provide a first step towards identifying microbial strains with favourable computational properties for future experimental implementations.

cs.ET↗

arXiv2010

Sensitivity analysis of a computational model of the IKK-NF-κB-IκBα-A20 signal transduction network

The NF-κB signaling network plays an important role in many different compartments of the immune system during immune activation. Using a computational model of the NF-κB signaling network involving two negative regulators, IκBα and A20, we performed sensitivity analyses with three different sampling methods and present a ranking of the kinetic rate variables by the strength of their influence on the NF-κB signaling response. We also present a classification of temporal response profiles of nuclear NF-κB concentration into six clusters, which can be regrouped to three biologically relevant clusters. Lastly, based upon the ranking, we constructed a reduced network of the IKK-NF-κB-IκBα-A20 signal transduction.

q-bio.QM↗

arXiv2010

Noise-induced oscillatory shuttling of NF-κB in a two compartment IKK-NF-κB-IκB-A20 signaling model

NF-κB is a pleiotropic protein whose nucleo-cytoplasmic trafficking is tightly regulated by multiple negative feedback loops embedded in the NF-κB signaling network and contributes to diverse gene expression profiles important in immune cell differentiation, cell apoptosis, and innate immunity. The intracellular signaling processes and their control mechanisms, however, are susceptible to both extrinsic and intrinsic noise. In this article, we present numerical evidence for a universal dynamic behavior of NF-κB, namely oscillatory nucleo-cytoplasmic shuttling, due to the fundamentally stochastic nature of the NF-κB signaling network. We simulated the effect of extrinsic noise with a deterministic ODE model, using a statistical ensemble approach, generating many copies of the signaling network with different kinetic rates sampled from a biologically feasible parameter space. We modeled the effect of intrinsic noise by simulating the same networks stochastically using the Gillespie algorithm. The results demonstrate that extrinsic noise diversifies the shuttling patterns of NF-κB response, whereas intrinsic noise induces oscillatory behavior in many of the otherwise non-oscillatory patterns. We identify two key model parameters which significantly affect the NF-κB dynamic response and deduce a two-dimensional phase-diagram of the NF-κB response as a function of these parameters. We conclude that if single-cell experiments are performed, a rich variety of NF-κB response will be observed, even if population-level experiments, which average response over large numbers of cells, do not evidence oscillatory behavior.

q-bio.MN↗

arXiv2010

Novel statistical ensemble analysis for simulating extrinsic noise-driven response in NF-κB signaling network

Cellular responses in the single cells are known to be highly heterogeneous and individualistic due to the strong influence by extrinsic and intrinsic noise. Here, we are concerned about how to model the extrinsic noise-induced heterogeneous response in the single cells under the constraints of experimentally obtained population-averaged response, but without much detailed kinetic information. We propose a novel statistical ensemble scheme where extrinsic noise is regarded as fluctuations in the values of kinetic parameters and such fluctuations are modeled by randomly sampling the kinetic rate constants from a uniform distribution. We consider a large number of signaling system replicates, each of which has the same network topology, but a uniquely different set of kinetic rate constants. A protein dynamic response from each replicate should represent the dynamics in a single cell and the statistical ensemble average should be regarded as a population-level response averaged over a population of the cells. We devise an optimization algorithm to find the correct uniform distribution of the network parameters, which produces the correct statistical distribution of the response whose ensemble average and distribution agree well with the population-level experimental data and the experimentally observed heterogeneity. We apply this statistical ensemble analysis to a NF-κB signaling system and (1) predict the distributions of the heterogeneous NF-κB (either oscillatory or non-oscillatory) dynamic patterns and of the dynamic features (e.g., period), (2) predict that both the distribution and the statistical ensemble average of the NF-κB dynamic response depends sensitively on the dosage of stimulant, and lastly (3) demonstrate the sigmoidally shaped dose-response from the statistical ensemble average and the individual replicates.

q-bio.MN↗