A Molecular Communication Scheme to Estimate the State of Biochemical Processes on a Lab-on-a-Chip

Bayram Cevdet Akdeniz, Malcolm EganNanoCoCoA@SenSys 2020 · DOI

TL;DR

This paper proposes a novel molecular communication scheme to estimate the equilibrium state of biochemical processes within a lab-on-a-chip (LoC) device. By leveraging the statistical properties of molecules diffusing through microfluidic channels and detected in a separate chamber, the system can reliably infer the biochemical state, even when the possible states are unknown beforehand. This opens doors for more sophisticated in-situ monitoring of complex biological systems.

What Problem Does This Paper Address?

Understanding the dynamics of biochemical processes is crucial in many fields, from drug discovery to synthetic biology. Traditionally, this involves performing assays that often require specific environmental conditions or invasive measurements. This can be problematic when trying to observe processes in their natural state (in situ or in vivo). An alternative is to use advanced spectroscopy methods, but these too can have limitations, such as requiring specialized substrates or potentially perturbing the very process being studied.

The authors address this by proposing a distributed sensing approach using lab-on-a-chip (LoC) technology. The core challenge is to reliably communicate information about the biochemical process's equilibrium state from where it occurs to a separate detection chamber, using microfluidic channels as the communication medium. This communication needs to be robust and accurate, even when the exact nature or number of possible equilibrium states of the biochemical process is not fully known.

Methods

The paper introduces a molecular communication scheme that relies on the diffusion of signaling molecules through microfluidic channels. The process involves:

  • Biochemical Process Chamber: Where the biochemical reaction of interest takes place and reaches an equilibrium state.
  • Microfluidic Channel: Acts as the communication channel. Molecules from the process chamber diffuse through this channel.
  • Spectroscopy Chamber: A separate chamber equipped with a passive spectroscopy device that counts the number of specific molecules (S3) that are produced from the signaling molecules (S2) diffusing from the channel.

The key innovation lies in how the information is encoded and decoded:

  • Signaling: Molecules (S1) from the biochemical process react to form signaling molecules (S2). These S2 molecules then diffuse through the microfluidic channel and react in the spectroscopy chamber to form detection molecules (S3).
  • Statistical Modeling: The number of S3 molecules detected in the spectroscopy chamber is modeled as a Gaussian distribution, conditioned on the specific equilibrium state of the biochemical process. This is based on the steady-state statistics of the reaction-diffusion master equation (RDME) governing the molecular dynamics.
  • Equilibrium State Estimation: When the number of equilibrium states and their properties are unknown, the authors propose using a Gaussian Mixture Model (GMM). The GMM is trained using measurements from multiple experiments (where the biochemical process is subjected to different environmental conditions). The expectation-maximization algorithm is used to estimate the parameters of the GMM, including the mean (representing the equilibrium state), variance, and mixing coefficients (representing the likelihood of each state).
  • Model Selection: To determine the optimal number of equilibrium states (clusters) for the GMM, the Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC) is employed.

Datasets

Not discussed in the paper. The evaluation is performed using simulations based on a model of a self-activating toggle switch circuit.

Key Insights and Ideas

  • Distributed Sensing via Molecular Communication: The paper demonstrates that microfluidic channels can effectively act as communication links for transmitting information about biochemical states, enabling distributed sensing architectures.
  • Statistical Inference from Molecular Counts: The core idea is that the statistical distribution of detected molecules (S3) in the spectroscopy chamber provides a reliable signature of the upstream biochemical equilibrium state. This leverages the inherent stochasticity of molecular processes.
  • Gaussian Mixture Models for Unknown States: The application of GMMs is a powerful approach for inferring the number and characteristics of equilibrium states when they are not pre-defined or fully characterized, making the system adaptable to novel biological systems.
  • Trade-offs in Measurement: The work highlights the trade-off between the rate of spectroscopy measurements and the accuracy of state estimation, providing practical guidelines for designing high-throughput biological assays.

Key Results

  • The proposed scheme can reliably estimate the equilibrium state of a biochemical process.
  • The accuracy of estimation improves with an increased number of samples ($N_s$) and a longer observation time ($T_s$).
  • For a simulated self-activating toggle switch circuit, which exhibits tristability (three equilibrium states), the GMM-based estimation algorithm successfully identified and differentiated these states.
  • The paper quantifies the detection error and estimation error (NMSE) as a function of the number of samples, demonstrating the expected decrease in error with more data.

Limitations and Future Work

  • Assumptions: The model assumes unimolecular reactions and diffusion-dominated dynamics in the microfluidic channel, without convection. The passive nature of the spectroscopy device is also a key assumption.
  • Parameter Sensitivity: The performance of the estimation relies on accurate knowledge or estimation of parameters like diffusion rates and reaction constants.
  • Future Work: The authors suggest extending the approach to systems with more complex reaction networks and exploring the implementation of these schemes on actual lab-on-a-chip devices. Further investigation into the impact of different microfluidic geometries and the integration of more sophisticated detection mechanisms are also potential avenues.

Summary updated

BibTeX
@inproceedings{akdeniz2020molecular,
  author = {Bayram Cevdet Akdeniz and Malcolm Egan},
  title = {{A Molecular Communication Scheme to Estimate the State of Biochemical Processes on a Lab-on-a-Chip}},
  booktitle = {NanoCoCoA@SenSys},
  year = {2020},
  doi = {10.1145/3416006.3431272},
}
Figure