Optimal Design of a Molecular Recognizer: Molecular Recognition as a Bayesian Signal Detection Problem

Numerous biological functions-such as enzymatic catalysis, the immune response system, and the DNA-protein regulatory network-rely on the ability of molecules to specifically recognize target molecules within a large pool of similar competitors in a noisy biochemical environment. Using the basic framework of signal detection theory, we treat the molecular recognition process as a signal detection problem and examine its overall performance. Thus, we evaluate the optimal properties of a molecular recognizer in the presence of competition and noise. Our analysis reveals that the optimal design undergoes a “phase transition” as the structural properties of the molecules and interaction energies between them vary. In one phase, the recognizer should be complementary in structure to its target (like a lock and a key), while in the other, conformational changes upon binding, which often accompany molecular recognition, enhance recognition quality. Using this framework, the abundance of conformational changes may be explained as a result of increasing the fitness of the recognizer. Furthermore, this analysis may be used in future design of artificial signal processing devices based on biomolecules.

💡 Research Summary

The paper reframes molecular recognition—a cornerstone of enzymatic catalysis, immune surveillance, and gene regulation—as a Bayesian signal detection problem. By treating the recognizer as a receiver, the target molecule as the signal, and competing, structurally similar molecules as noise, the authors construct a quantitative framework that captures the stochastic nature of biochemical environments.

Key to the model are two design parameters of the recognizer: structural complementarity (the classic “lock‑and‑key” fit) and the propensity for conformational change upon binding (induced fit). The authors assign prior probabilities based on the relative concentrations of target and competitors, and they model binding free‑energy changes as likelihood functions. An asymmetric cost function penalizes false positives (binding a non‑target) and false negatives (missing the target) with different weights, reflecting biological selective pressures.

The expected utility, defined as the sum over all states of (probability ×