Biology subjects

Zaikin, A.

Publications and source records attributed to Zaikin, A..

7 recordsLinked to original sources

bioRxiv2026

Nonlinear regime transitions enable reservoir computation in activator-inhibitor cellular automata

Activator-inhibitor systems are usually studied as pattern-forming media, but the same local nonlinear interactions can also shape how information is stored and separated over time. Here we first introduce a reaction-diffusion-inspired cellular automaton as a tunable nonlinear medium with two activation families, a continuous sigmoid and a logistic-step relaxer. In each case, a single family-specific parameter changes the nonlinear response or relaxation dynamics while leaving the neighbourhood wiring fixed. This provides a controlled way to move the activator-inhibitor system between collapsed, structured, saturated, and overshooting regimes, and to relate these regimes to state-level diversity and pattern compressibility. We then use reservoir computing to test whether these tuned pattern-forming regimes support fading memory, input separability, and downstream readout learning. In this role, reservoir computing is not treated as an architecture to be optimised, but as a diagnostic of the computational properties of the untrained medium. Across parameter sweeps, stronger memory and separability are concentrated near transition regions rather than distributed uniformly across parameter space. These results identify activator-inhibitor cellular automata as interpretable unconventional reservoir substrates in which effective gain, threshold, and relaxation parameters tune the balance between pattern formation, memory, and separability. More broadly, they support the view that tissue-like or physical pattern-forming media may, in principle, shift between patterning and information-processing regimes by modulating local nonlinear dynamics.

systems biology↗

bioRxiv2023

Emergent digital bio-computation through spatial diffusion and engineered bacteria

Building computationally capable biological systems has long been an aim of synthetic biology. The potential utility of biocomputing devices ranges from biosafety and environmental applications to diagnosis and personalised medicine. Here we present work for the design of bacterial computers which use spatial patterning to process information. Our computers are composed of bacterial colonies which, inspired by patterning in embryo development, receive information in the form of diffusible morphogen-like signals. A computation is encoded by the physical locations of the input sources of morphogen and the output receiver colonies. We demonstrate, mathematically and with engineered Escherichia coli, the simple digital logic capability of single bacterial colonies and show how additional colonies are required to build complex functions. Inspired by electronic design automation, an algorithm for designing optimal spatial circuits computing two-level digital functions is presented. This enhances the capability of our system to complex digital functions without increasing the biological complexity. We extend our experimental system to incorporate biosensing colonies as morphogen sources, demonstrating how a diagnostic device might be constructed. Our approach will open up new ways to perform biological computation, with applications in bioengineering, biomaterials and biosensing. Ultimately, these computational bacterial communities will help us explore information processing in natural biological systems.

synthetic biology↗

bioRxiv2023

TIME-SERIES CLASSIFICATION FOR PATIENTS UNDER ACTIVE SURVEILLANCE AND SCREENING USING ECHO STATE NETWORK

AO_SCPLOWBSTRACTC_SCPLOWOver the past few decades, more and more patients come on follow-up studies such as active surveillance and screening, which results in a vast amount of time-series data in the health department. Each Patient typically has a small but different number of visits to the doctor and the time interval between the visits is heterogeneous. Nowadays, many machine learning tasks in relation to time series data are carried out using deep recurrent neural networks (RNN). However, deep neural networks consume enormous computational power as all weights in the network need to be trained through back-propagation. Conversely, echo state network (ESN), another form of RNN, demonstrates low training cost and the potential of it is still largely untapped. Therefore, in this article we will develop a new methodology that can classify aforementioned time-series data using the echo state network. We will also discuss how to address the heterogeneity in the time interval arising from the data of this type and how our model can also potentially fit other time-series data.

bioinformatics↗

bioRxiv2023

Technical and biological sources of unreliability of Infinium type II probes of the Illumina MethylationEPIC BeadChip microarray

The Illumina Methylation array platform has facilitated countless epigenetic studies on DNA methylation (DNAme) in health and disease, yet relatively few studies have so studied its reliability, i.e., the consistency of repeated measures. Here we focus on the reliability of both type I and type II Infinium probes. We propose a method for excluding unreliable probes based on dynamic thresholds for mean intensity (MI) and unreliability, estimated by probe-level simulation of the influence of technical noise on methylation {beta}-values using the background intensities of negative control probes. We validate our method in several datasets, including Illumina MethylationEPIC BeadChip v1.0 data from paired whole blood samples taken six weeks apart. Our analysis revealed that specifically probes with low MI exhibit higher {beta}-value variability between repeated samples. MI was associated with the number of C-bases in the respective probe sequence and correlated negatively with unreliability scores. The unreliability scores were substantiated through validation in a new EPIC v1.0 (blood and cervix) and a publicly available 450k (blood) dataset, as they effectively captured the variability observed in {beta}-values between technical replicates. Finally, despite promising higher robustness, the newer version v2.0 of the MethylationEPIC BeadChip retained a substantial number of probes with poor unreliability scores. To enhance current pre-processing pipelines, we developed an R package to calculate MI and unreliability scores and provide guidance on establishing optimal dynamic score thresholds for a given data set.

genomics↗

bioRxiv2020

Disentangling age-dependent DNA methylation: deterministic, stochastic, and nonlinear

DNA methylation variability arises due to concurrent genetic and environmental influences. Each of them is a mixture of regular and noisy sources, whose relative contribution has not been satisfactorily understood yet. We conduct a systematic assessment of the age-dependent methylation by the signal-to-noise ratio and identify a wealth of "deterministic" CpG probes (about 90%), whose methylation variability likely originates due to genetic and general environmental factors. The remaining 10% of "stochastic" CpG probes are arguably governed by the biological noise or incidental environmental factors. Investigating the mathematical functional relationship between methylation levels and variability, we find that in about 90% of the age-associated differentially methylated positions, the variability changes as the square of the methylation level, whereas in the most of the remaining cases the dependence is linear. Furthermore, we demonstrate that the methylation level itself in more than 15% cases varies nonlinearly with age (according to the power law), in contrast to the previously assumed linear changes. Our findings present ample evidence of the ubiquity of strong DNA methylation regulation, resulting in the individual age-dependent and nonlinear methylation trajectories, whose divergence explains the cross-sectional variability. It may also serve a basis for constructing novel nonlinear epigenetic clocks.

systems biology↗

bioRxiv2020

Age-dependent DNA methylation Parenclitic Networks in family-based cohort patients with Down Syndrome

Despite the fact that the cause of Down Syndrome (DS) is well established, the underlying molecular mechanisms that contribute to the syndrome and the phenotype of accelerated aging remain largely unknown. DNA methylation profiles are largely altered in DS, but it remains unclear how different methylation regions and probes are structured into a network of interactions. We develop and generalize the Parenclitic Networks approach that enables finding correlations between distant CpG probes (which are not pronounced as stand-alone biomarkers) and quantifies hidden network changes in DNA methylation. DS and a familybased cohort (including healthy siblings and mothers of persons with DS) are used as a case study. Following this approach, we constructed parenclitic networks and obtained different signatures that indicate (i) differences between individuals with DS and healthy individuals; (ii) differences between young and old healthy individuals; (iii) differences between DS individuals and their age-matched siblings, and (iv) difference between DS and the adult population (their mothers). The Gene Ontology analysis showed that the CpG network approach is more powerful than the single CpG approach in identifying biological processes related to DS phenotype. This includes the processes occurring in the central nervous system, skeletal muscles, disorders in carbohydrate metabolism, cardiopathology, and oncogenes. Our open-source software implementation is accessible to all researchers. The software includes a complete workflow, which can be used to construct Parenclitic Networks with any machine learning algorithm as a kernel to build edges. We anticipate a broad applicability of the approach to other diseases.

bioinformatics↗

bioRxiv2018

Impact of mitochondrial epistatic interactions on evolution of different human subpopulations

Investigation of human mitochondrial (mt) genome variation has been shown to provide insights to the human history and natural selection. By analyzing 24,167 human mt-genome samples, collected for five continents, we have developed a co-mutation network model to investigate characteristic human evolutionary patterns. The analysis highlighted richer co-mutating regions of the mt-genome, suggesting the presence of epistasis. Specifically, a large portion of COX genes was found to co-mutate in Asian and American populations, whereas, in African, European, and Oceanic populations, there was greater co-mutation bias in hypervariable regions. Interestingly, this study demonstrated hierarchical modularity as a crucial agent for these co-mutation networks. More profoundly, our ancestry-based co-mutation module analyses showed that mutations cluster preferentially in known mitochondrial haplogroups. Contemporary human mt-genome nucleotides most closely resembled the ancestral state, and very few of them were found to be ancestral-variants. Overall, these results demonstrated that subpopulation-based biases may favor mitochondrial gene specific epistasis.

synthetic biology↗