Bacteria evolved genes whose single-cell distributions of expression levels are broad, or even bimodal. Evidence suggests that they might enhance phenotypic diversity for coping with fluctuating environments. We identified seven genes in E. coli with bimodal (low and high) single-cell expression levels under standard growth conditions and studied how their dynamics are modified by environmental and antibiotic stresses known to target gene expression. We found that all genes lose bimodality under some, but not under all, stresses. Also, bimodality can reemerge upon cells returning to standard conditions, which suggests that the genes can switch often between high and low expression rates. As such, these genes could become valuable components of future multi-stable synthetic circuits. Next, we proposed models of bimodal transcription dynamics with realistic parameter values, able to mimic the outcome of the perturbations studied. We explored several models' tunability and boundaries of parameter values, beyond which it shifts to unimodal dynamics. From the model results, we predict that bimodality is robust, and yet tunable, not only by RNA and protein degradation rates, but also by the fraction of time that promoters remain unavailable for new transcription events. Finally, we show evidence that, although the empirical expression levels are influenced by many factors, the bimodality emerges during transcription initiation, at the promoter regions and, thus, may be evolvable and adaptable.

Reinforcement learning algorithms have been shown to generate procedures for executing quantum control tasks to desired performances. Although reinforcement learning has been effective, its robustness in generating quantum control procedures under general noise condition needs to be tested. Here we consider adaptive quantum-enhanced interferometric phase estimation as a case study in quantum feedback control. The algorithm is determined to deliver a robust quantum-metrological procedure under various phase-noise models when the imprecision surpasses the standard quantum limit. The robustness against unknown environmental variations positions the reinforcement learning as a practical approach for devising quantum control procedures.

The concept of generalized concatenated quantum codes (GCQC) provides a systematic way for constructing good quantum codes from short component codes. We introduce a stabilizer formalism for GCQCs, which is achieved by defining quantum coset codes. This formalism offers a new perspective for GCQCs and enables us to derive a lower bound on the code distance of stabilizer GCQCs from component codes parameters, for both non-degenerate and degenerate component codes. Our formalism also shows how to exploit the error-correcting capacity of component codes to design good GCQCs efficiently.

Decoding sparse quantum codes can be accomplished by syndrome-based decoding using a belief propagation (BP) algorithm. We significantly improve this decoding scheme by developing a new feedback adjustment strategy for the standard BP algorithm. In our feedback procedure, we exploit much of the information from stabilizers, not just the syndrome but also the values of the frustrated checks on individual qubits of the code and the channel model. Furthermore we show that our decoding algorithm is superior to belief propagation algorithms using only the syndrome in the feedback procedure for all cases of the depolarizing channel. Our algorithm does not increase the measurement overhead compared to the previous method, as the extra information comes for free from the requisite stabilizer measurements.