Photonic Stochastic Emergent Storage for deep classification by scattering-intrinsic patterns.

Disorder is a pervasive characteristic of natural systems, offering a wealth of non-repeating patterns. In this study, we present a novel storage method that harnesses naturally-occurring random structures to store an arbitrary pattern in a memory device. This method, the Stochastic Emergent Storage (SES), builds upon the concept of emergent archetypes, where a training set of imperfect examples (prototypes) is employed to instantiate an archetype in a Hopfield-like network through emergent processes. We demonstrate this non-Hebbian paradigm in the photonic domain by utilizing random transmission matrices, which govern light scattering in a white-paint turbid medium, as prototypes. Through the implementation of programmable hardware, we successfully realize and experimentally validate the capability to store an arbitrary archetype and perform classification at the speed of light. Leveraging the vast number of modes excited by mesoscopic diffusion, our approach enables the simultaneous storage of thousands of memories without requiring any additional fabrication efforts. Similar to a content addressable memory, all stored memories can be collectively assessed against a given pattern to identify the matching element. Furthermore, by organizing memories spatially into distinct classes, they become features within a higher-level categorical (deeper) optical classification layer.

A recurrent Hopfield network for estimating meso-scale effective connectivity in MEG.

The architecture of communication within the brain, represented by the human connectome, has gained a paramount role in the neuroscience community. Several features of this communication, e.g., the frequency content, spatial topology, and temporal dynamics are currently well established. However, identifying generative models providing the underlying patterns of inhibition/excitation is very challenging. To address this issue, we present a novel generative model to estimate large-scale effective connectivity from MEG. The dynamic evolution of this model is determined by a recurrent Hopfield neural network with asymmetric connections, and thus denoted Recurrent Hopfield Mass Model (RHoMM). Since RHoMM must be applied to binary neurons, it is suitable for analyzing Band Limited Power (BLP) dynamics following a binarization process. We trained RHoMM to predict the MEG dynamics through a gradient descent minimization and we validated it in two steps. First, we showed a significant agreement between the similarity of the effective connectivity patterns and that of the interregional BLP correlation, demonstrating RHoMM's ability to capture individual variability of BLP dynamics. Second, we showed that the simulated BLP correlation connectomes, obtained from RHoMM evolutions of BLP, preserved some important topological features, e.g, the centrality of the real data, assuring the reliability of RHoMM. Compared to other biophysical models, RHoMM is based on recurrent Hopfield neural networks, thus, it has the advantage of being data-driven, less demanding in terms of hyperparameters and scalable to encompass large-scale system interactions. These features are promising for investigating the dynamics of inhibition/excitation at different spatial scales.

Reference-less wavefront shaping in a Hopfield-like rough intensity landscape.

This study introduces a new digital-micromirror based binary-phase wavefront shaping technique, which allows the measurement of the full coupling matrix of a disordered medium without a reference and enables to focusing transmitted light. The coupling matrix takes on a bi-dyadic structure, similar to a Hopfield memory matrix containing two memory patterns. Sequential wavefront optimization in this configuration often stalls due to a rough intensity landscape, resulting in a non-optimal state. To overcome this issue, we propose the Complete Couplings Mapping method, which consistently reaches the theoretically expected maximum intensity.

Collective behavior and self-organization in neural rosette morphogenesis.

Neural rosettes develop from the self-organization of differentiating human pluripotent stem cells. This process mimics the emergence of the embryonic central nervous system primordium, i.e., the neural tube, whose formation is under close investigation as errors during such process result in severe diseases like spina bifida and anencephaly. While neural tube formation is recognized as an example of self-organization, we still do not understand the fundamental mechanisms guiding the process. Here, we discuss the different theoretical frameworks that have been proposed to explain self-organization in morphogenesis. We show that an explanation based exclusively on stem cell differentiation cannot describe the emergence of spatial organization, and an explanation based on patterning models cannot explain how different groups of cells can collectively migrate and produce the mechanical transformations required to generate the neural tube. We conclude that neural rosette development is a relevant experimental 2D in-vitro model of morphogenesis because it is a multi-scale self-organization process that involves both cell differentiation and tissue development. Ultimately, to understand rosette formation, we first need to fully understand the complex interplay between growth, migration, cytoarchitecture organization, and cell type evolution.

Differences in the organization of interface residues tunes the stability of the SARS-CoV-2 spike-ACE2 complex.

The continuous emergence of novel variants represents one of the major problems in dealing with the SARS-CoV-2 virus. Indeed, also due to its prolonged circulation, more than ten variants of concern emerged, each time rapidly overgrowing the current viral version due to improved spreading features. As, up to now, all variants carry at least one mutation on the spike Receptor Binding Domain, the stability of the binding between the SARS-CoV-2 spike protein and the human ACE2 receptor seems one of the molecular determinants behind the viral spreading potential. In this framework, a better understanding of the interplay between spike mutations and complex stability can help to assess the impact of novel variants. Here, we characterize the peculiarities of the most representative variants of concern in terms of the molecular interactions taking place between the residues of the spike RBD and those of the ACE2 receptor. To do so, we performed molecular dynamics simulations of the RBD-ACE2 complexes of the seven variants of concern in comparison with a large set of complexes with different single mutations taking place on the RBD solvent-exposed residues and for which the experimental binding affinity was available. Analyzing the strength and spatial organization of the intermolecular interactions of the binding region residues, we found that (i) mutations producing an increase of the complex stability mainly rely on instaurating more favorable van der Waals optimization at the cost of Coulombic ones. In particular, (ii) an anti-correlation is observed between the shape and electrostatic complementarities of the binding regions. Finally, (iii) we showed that combining a set of dynamical descriptors is possible to estimate the outcome of point mutations on the complex binding region with a performance of 0.7. Overall, our results introduce a set of dynamical observables that can be rapidly evaluated to probe the effects of novel isolated variants or different molecular systems.

Electrostatic complementarity at the interface drives transient protein-protein interactions.

Understanding the mechanisms driving bio-molecules binding and determining the resulting complexes' stability is fundamental for the prediction of binding regions, which is the starting point for drug-ability and design. Characteristics like the preferentially hydrophobic composition of the binding interfaces, the role of van der Waals interactions, and the consequent shape complementarity between the interacting molecular surfaces are well established. However, no consensus has yet been reached on the role of electrostatic. Here, we perform extensive analyses on a large dataset of protein complexes for which both experimental binding affinity and pH data were available. Probing the amino acid composition, the disposition of the charges, and the electrostatic potential they generated on the protein molecular surfaces, we found that (i) although different classes of dimers do not present marked differences in the amino acid composition and charges disposition in the binding region, (ii) homodimers with identical binding region show higher electrostatic compatibility with respect to both homodimers with non-identical binding region and heterodimers. Interestingly, (iii) shape and electrostatic complementarity, for patches defined on short-range interactions, behave oppositely when one stratifies the complexes by their binding affinity: complexes with higher binding affinity present high values of shape complementarity (the role of the Lennard-Jones potential predominates) while electrostatic tends to be randomly distributed. Conversely, complexes with low values of binding affinity exploit Coulombic complementarity to acquire specificity, suggesting that electrostatic complementarity may play a greater role in transient (or less stable) complexes. In light of these results, (iv) we provide a novel, fast, and efficient method, based on the 2D Zernike polynomial formalism, to measure electrostatic complementarity without the need of knowing the complex structure. Expanding the electrostatic potential on a basis of 2D orthogonal polynomials, we can discriminate between transient and permanent protein complexes with an AUC of the ROC of [Formula: see text] 0.8. Ultimately, our work helps shedding light on the non-trivial relationship between the hydrophobic and electrostatic contributions in the binding interfaces, thus favoring the development of new predictive methods for binding affinity characterization.

Hedgehog-GLI and Notch Pathways Sustain Chemoresistance and Invasiveness in Colorectal Cancer and Their Inhibition Restores Chemotherapy Efficacy.

Colorectal cancer (CRC) is a leading cause of cancer-related mortality and chemoresistance is a major medical issue. The epithelial-to-mesenchymal transition (EMT) is the primary step in the emergence of the invasive phenotype and the Hedgehog-GLI (HH-GLI) and NOTCH signaling pathways are associated with poor prognosis and EMT in CRC. CRC cell lines harboring KRAS or BRAF mutations, grown as monolayers and organoids, were treated with the chemotherapeutic agent 5-Fluorouracil (5-FU) alone or combined with HH-GLI and NOTCH pathway inhibitors GANT61 and DAPT, or arsenic trioxide (ATO) to inhibit both pathways. Treatment with 5-FU led to the activation of HH-GLI and NOTCH pathways in both models. In KRAS mutant CRC, HH-GLI and NOTCH signaling activation co-operate to enhance chemoresistance and cell motility, while in BRAF mutant CRC, the HH-GLI pathway drives the chemoresistant and motile phenotype. We then showed that 5-FU promotes the mesenchymal and thus invasive phenotype in KRAS and BRAF mutant organoids and that chemosensitivity could be restored by targeting the HH-GLI pathway in BRAF mutant CRC or both HH-GLI and NOTCH pathways in KRAS mutant CRC. We suggest that in KRAS-driven CRC, the FDA-approved ATO acts as a chemotherapeutic sensitizer, whereas GANT61 is a promising chemotherapeutic sensitizer in BRAF-driven CRC.

A low-cost, label-free microfluidic scanning flow cytometer for high-accuracy quantification of size and refractive index of particles.

Flow cytometers and fluorescence activated cells sorters (FCM/FACS) represent the gold standard for high-throughput single-cell analysis, but their usefulness for label-free applications is limited by the unreliability of forward and side scatter measurements. Scanning flow cytometers represent an appealing alternative, as they exploit measurements of the angle-resolved scattered light to provide accurate and quantitative estimates of cellular properties, but the requirements of current setups are unsuitable for integration with other lab-on-chip technologies or for point-of-care applications. Here we present the first microfluidic scanning flow cytometer (µSFC), able to achieve accurate angle-resolved scattering measurements within a standard polydimethylsiloxane microfluidic chip. The system exploits a low cost linearly variable optical density (OD) filter to reduce the dynamic range of the signal and to increase its signal-to-noise ratio. We present a performance comparison between the µSFC and commercial machines for the label free characterization of polymeric beads with different diameters and refractive indices. In contrast to FCM and FACS, the µSFC yields size estimates linearly correlated with nominal particle sizes (R2 = 0.99) and quantitative estimates of particle refractive indices. The feasibility of using the µSFC for the characterization of biological samples is demonstrated by analyzing a population of monocytes identified based on the morphology of a peripheral blood mononuclear cells sample, which yields values in agreement with the literature. The proposed µSFC combines low setup requirements with high performance, and has great potential for integration within other lab-on-chip systems for multi-parametric cell analysis and for next-generation point-of-care diagnostic applications.

Physics-informed deep neural network for image denoising.

Image enhancement deep neural networks (DNN) can improve signal to noise ratio or resolution of optically collected visual information. The literature reports a variety of approaches with varying effectiveness. All these algorithms rely on arbitrary data (the pixels' count-rate) normalization, making their performance strngly affected by dataset or user-specific data pre-manipulation. We developed a DNN algorithm capable to enhance images signal-to-noise surpassing previous algorithms. Our model stems from the nature of the photon detection process which is characterized by an inherently Poissonian statistics. Our algorithm is thus driven by distance between probability functions instead than relying on the sole count-rate, producing high performance results especially in high-dynamic-range images. Moreover, it does not require any arbitrary image renormalization other than the transformation of the camera's count-rate into photon-number.

Deep learning for blind structured illumination microscopy.

Blind-structured illumination microscopy (blind-SIM) enhances the optical resolution without the requirement of nonlinear effects or pre-defined illumination patterns. It is thus advantageous in experimental conditions where toxicity or biological fluctuations are an issue. In this work, we introduce a custom convolutional neural network architecture for blind-SIM: BS-CNN. We show that BS-CNN outperforms other blind-SIM deconvolution algorithms providing a resolution improvement of 2.17 together with a very high Fidelity (artifacts reduction). Furthermore, BS-CNN proves to be robust in cross-database variability: it is trained on synthetically augmented open-source data and evaluated on experiments. This approach paves the way to the employment of CNN-based deconvolution in all scenarios in which a statistical model for the illumination is available while the specific realizations are unknown or noisy.

Spatial coherence of light inside three-dimensional media.

Speckle is maybe the most fundamental interference effect of light in disordered media, giving rise to fascinating physical phenomena and cutting edge applications. While speckle formed outside a sample is easily measured and analysed, true bulk speckle, as formed inside random media, is difficult to investigate directly due to the obvious issue of physical access. Furthermore, its proper theoretical description poses enormous challenges. Here we report on the first direct measurements of spatially resolved intensity correlations of light inside a disordered medium, using embedded DNA strings decorated with emitters separated by a controlled nanometric distance. Our method provides in situ access to fundamental properties of bulk speckles as their size and polarization degrees of freedom, both of which are found to deviate significantly from theoretical predictions. The deviations are explained, by comparison with rigorous numerical calculations, in terms of correlations among polarization components and non-universal near-field contributions at the nanoscale.

Optonongenetic enhancement of activity in primary cortical neurons.

It has been recently demonstrated that the exposure of naive neuronal cells to light-at the basis of optogenetic techniques and calcium imaging measurements-may alter neuronal firing. Indeed, understanding the effect of light on nongenetically modified neurons is crucial for a correct interpretation of calcium imaging and optogenetic experiments. Here we investigated the effect of continuous visible LED light exposure (490 nm, $ 0.18 {-} 1.3\;{\rm mW}/{{\rm mm}^2} $0.18-1.3mW/mm2) on spontaneous activity of primary neuronal networks derived from the early postnatal mouse cortex. We demonstrated, by calcium imaging and patch clamp experiments, that illumination higher than $ 1.0\;{\rm mW}/{{\rm mm}^2} $1.0mW/mm2 causes an enhancement of network activity in cortical cultures. We investigated the possible origin of the phenomena by blocking the transient receptor potential vanilloid 4 (TRPV4) channel, demonstrating a complex connection between this temperature-dependent channel and the measured effect. The results presented here shed light on an exogenous artifact, potentially present in all calcium imaging experiments, that should be taken into account in the analysis of fluorescence data.

Localization-based two-photon wave-function information encoding.

In quantum communications, quantum states are employed for the transmission of information between remote parties. This usually requires sharing knowledge of the measurement bases through a classical public channel in the sifting phase of the protocol. Here, we demonstrate a quantum communication scheme where the information on the bases is shared "non-classically," by encoding this information in the same photons used for carrying the data. This enhanced capability is achieved by exploiting the localization of the photonic wave function, observed when the photons are prepared and measured in the same quantum basis. We experimentally implement our scheme by using a multi-mode optical fiber coupled to an adaptive optics setup. We observe a decrease in the error rate for higher dimensionality, indicating an improved resilience against noise.

Scattering Assisted Imaging.

Standard imaging systems provide a spatial resolution that is ultimately dictated by the numerical aperture (NA) of the illumination and collection optics. In biological tissues, the resolution is strongly affected by scattering, which limits the penetration depth to a few tenths of microns. Here, we exploit the properties of speckle patterns embedded into a strongly scattering matrix to illuminate the sample at high spatial frequency content. Combining adaptive optics with a custom deconvolution algorithm, we obtain an increase in the transverse spatial resolution by a factor of 2.5 with respect to the natural diffraction limit. Our Scattering Assisted Imaging (SAI) provides an effective solution to increase the resolution when long working distance optics are needed, potentially paving the way to bulk imaging in turbid tissues.

Beyond the Maximum Storage Capacity Limit in Hopfield Recurrent Neural Networks.

In a neural network, an autapse is a particular kind of synapse that links a neuron onto itself. Autapses are almost always not allowed neither in artificial nor in biological neural networks. Moreover, redundant or similar stored states tend to interact destructively. This paper shows how autapses together with stable state redundancy can improve the storage capacity of a recurrent neural network. Recent research shows how, in an N-node Hopfield neural network with autapses, the number of stored patterns (P) is not limited to the well known bound 0.14 N , as it is for networks without autapses. More precisely, it describes how, as the number of stored patterns increases well over the 0.14 N threshold, for P much greater than N, the retrieval error asymptotically approaches a value below the unit. Consequently, the reduction of retrieval errors allows a number of stored memories, which largely exceeds what was previously considered possible. Unfortunately, soon after, new results showed that, in the thermodynamic limit, given a network with autapses in this high-storage regime, the basin of attraction of the stored memories shrinks to a single state. This means that, for each stable state associated with a stored memory, even a single bit error in the initial pattern would lead the system to a stationary state associated with a different memory state. This thus limits the potential use of this kind of Hopfield network as an associative memory. This paper presents a strategy to overcome this limitation by improving the error correcting characteristics of the Hopfield neural network. The proposed strategy allows us to form what we call an absorbing-neighborhood of state surrounding each stored memory. An absorbing-neighborhood is a set defined by a Hamming distance surrounding a network state, which is an absorbing because, in the long-time limit, states inside it are absorbed by stable states in the set. We show that this strategy allows the network to store an exponential number of memory patterns, each surrounded with an absorbing-neighborhood with an exponentially growing size.

Hyperuniformity in amorphous speckle patterns.

Hyperuniform structures possess the ability to confine and drive light, although their fabrication is extremely challenging. Here we demonstrate that speckle patterns obtained by a superposition of randomly arranged sources of Bessel beams can be used to generate hyperunifrom scalar fields. By exploiting laser light tailored with a spatial filter, we experimentally produce (without requiring any computational power) a speckle pattern possessing maxima at locations corresponding to a hyperuniform distribution. By properly filtering out intensity fluctuation from the same speckle pattern, it is possible to retrieve an intensity profile satisfying the hyperuniformity requirements. Our findings are supported by extensive numerical simulations.

Effect of dilution in asymmetric recurrent neural networks.

We study with numerical simulation the possible limit behaviors of synchronous discrete-time deterministic recurrent neural networks composed of N binary neurons as a function of a network's level of dilution and asymmetry. The network dilution measures the fraction of neuron couples that are connected, and the network asymmetry measures to what extent the underlying connectivity matrix is asymmetric. For each given neural network, we study the dynamical evolution of all the different initial conditions, thus characterizing the full dynamical landscape without imposing any learning rule. Because of the deterministic dynamics, each trajectory converges to an attractor, that can be either a fixed point or a limit cycle. These attractors form the set of all the possible limit behaviors of the neural network. For each network we then determine the convergence times, the limit cycles' length, the number of attractors, and the sizes of the attractors' basin. We show that there are two network structures that maximize the number of possible limit behaviors. The first optimal network structure is fully-connected and symmetric. On the contrary, the second optimal network structure is highly sparse and asymmetric. The latter optimal is similar to what observed in different biological neuronal circuits. These observations lead us to hypothesize that independently from any given learning model, an efficient and effective biologic network that stores a number of limit behaviors close to its maximum capacity tends to develop a connectivity structure similar to one of the optimal networks we found.

What is the Right Theory for Anderson Localization of Light? An Experimental Test.

Anderson localization of light is traditionally described in analogy to electrons in a random potential. Within this description, the random potential depends on the wavelength of the incident light. For transverse Anderson localization, this leads to the prediction that the distribution of localization lengths-and, hence, its average-strongly depends on the wavelength. In an alternative description, in terms of a spatially fluctuating electric modulus, this is not the case. Here, we report on an experimentum crucis in order to investigate the validity of the two conflicting theories using optical samples exhibiting transverse Anderson localization. We do not find any dependence of the observed average localization radii on the light wavelength. We conclude that the modulus-type description is the correct one and not the potential-type one. We corroborate this by showing that in the derivation of the traditional potential-type theory, a term in the wave equation has been tacitly neglected. In our new modulus-type theory, the wave equation is exact. We check the consistency of the new theory with our data using the nonlinear sigma model. We comment on the consequences for the general case of three-dimensional disorder.

Disorder-induced single-mode transmission.

Localized states trap waves propagating in a disordered potential and play a crucial role in Anderson localization, which is the absence of diffusion due to disorder. Some localized states are barely coupled with neighbours because of differences in wavelength or small spatial overlap, thus preventing energy leakage to the surroundings. This is the same degree of isolation found in the homogeneous core of a single-mode optical fibre. Here we show that localized states of a disordered optical fibre are single mode: the transmission channels possess a high degree of resilience to perturbation and invariance with respect to the launch conditions. Our experimental approach allows identification and characterization of the single-mode transmission channels in a disordered matrix, demonstrating low losses and densely packed single modes. These disordered and wavelength-sensitive channels may be exploited to de-multiplex different colours at different locations.

Secure information transport by transverse localization of light.

A single-photon beating with itself can produce even the most elaborate optical fringe pattern. However, the large amount of information enclosed in such a pattern is typically inaccessible, since the complete distribution can be visualized only after many detections. In fact this limitation is only true for delocalized patterns. Here we demonstrate how reconfigurable localized optical patterns allow to encode up to 6 bits of information in disorder-induced high transmission channels, even using a small number of photon counts. We developed a quantum key distribution scheme for fiber communication in which high information capacity is achieved through position and momentum complementarity.