A C C Coolen

The potential of optical proteomic technologies to individualize prognosis and guide rational treatment for cancer patients

Genomics and proteomics will improve outcome prediction in cancer and have great potential to help in the discovery of unknown mechanisms of metastasis, ripe for therapeutic exploitation. Current methods of prognosis estimation rely on clinical data, anatomical staging and histopathological features. It is hoped that translational genomic and proteomic research will discriminate more accurately than is possible at present between patients with a good prognosis and those who carry a high risk of recurrence. Rational treatments, targeted to the specific molecular pathways of an individual’s high-risk tumor, are at the core of tailored therapy. The aim of targeted oncology is to select the right patient for the right drug at precisely the right point in their cancer journey. Optical proteomics uses advanced optical imaging technologies to quantify the activity states of and associations between signaling proteins by measuring energy transfer between fluorophores attached to specific proteins. Förster resonance energy transfer (FRET) and fluorescence lifetime imaging microscopy (FLIM) assays are suitable for use in cell line models of cancer, fresh human tissues and formalin-fixed paraffin-embedded tissue (FFPE). In animal models, dynamic deep tissue FLIM/FRET imaging of cancer cells in vivo is now also feasible. Analysis of protein expression and post-translational modifications such as phosphorylation and ubiquitination can be performed in cell lines and are remarkably efficiently in cancer tissue samples using tissue microarrays (TMAs). FRET assays can be performed to quantify protein-protein interactions within FFPE tissue, far beyond the spatial resolution conventionally associated with light or confocal laser microscopy. Multivariate optical parameters can be correlated with disease relapse for individual patients. FRET-FLIM assays allow rapid screening of target modifiers using high content drug screens. Specific protein-protein interactions conferring a poor prognosis identified by high content tissue screening will be perturbed with targeted therapeutics. Future targeted drugs will be identified using high content/throughput drug screens that are based on multivariate proteomic assays. Response to therapy at a molecular level can be monitored using these assays while the patient receives treatment: utilizing re-biopsy tumor tissue samples in the neoadjuvant setting or by examining surrogate tissues. These technologies will prove to be both prognostic of risk for individuals when applied to tumor tissue at first diagnosis and predictive of response to specifically selected targeted anticancer drugs. Advanced optical assays have great potential to be translated into real-life benefit for cancer patients.

How Forster Resonance Energy Transfer Imaging Improves the Understanding of Protein Interaction Networks in Cancer Biology

Herein we discuss how FRET imaging can contribute at various stages to delineate the function of the proteome. Therefore, we briefly describe FRET imaging techniques, the selection of suitable FRET pairs and potential caveats. Furthermore, we discuss state-of-the-art FRET-based screening approaches (underpinned by protein interaction network analysis using computational biology) and preclinical intravital FRET-imaging techniques that can be used for functional validation of candidate hits (nodes and edges) from the network screen, as well as measurement of the efficacy of perturbing these nodes/edges by short hairpin RNA (shRNA) and/or small molecule-based approaches.

Phase Diagram and Storage Capacity of Sequence-Storing Neural Networks

We solve the dynamics of Hopfield-type neural networks which store sequences of patterns, close to saturation. The asymmetry of the interaction matrix in such models leads to violation of detailed balance, ruling out an equilibrium statistical mechanical analysis. Using generating functional methods we derive exact closed equations for dynamical order parameters, viz. the sequence overlap and correlation and response functions. in the limit of an infinite system size. We calculate the time translation invariant solutions of these equations. describing stationary limit-cycles. which leads to a phase diagram. The effective retarded self-interaction usually appearing in symmetric models is here found to vanish, which causes a significantly enlarged storage capacity of αc ≈ 0.269. compared to αc ≈ 0.139 for Hopfield networks storing static patterns. Our results are tested against extensive computer simulations and excellent agreement is found.

Integrating Receptor Signal Inputs That Influence Small Rho GTPase Activation Dynamics at the Immunological Synapse

ABSTRACT The Rho GTPase Cdc42 regulates cytoskeletal changes at the immunological synapse (IS) that are critical to T-cell activation. By imaging fluorescent activity biosensors (Raichu) using fluorescence lifetime imaging microscopy, Cdc42 activation was shown to display kinetics that are conditional on the specific receptor input (through two IS-associated receptors, CD3 and β1 integrin). CD3-triggered Cdc42 activity is dependent on the cyto-2 (NPIY) motif of the β1 integrin cytoplasmic domain. Perturbations of the ezrin-radixin-moesin (ERM) function blocked CD3- and β1-dependent increases in Cdc42 activity. Both IS-associated receptors probably lie on a serial molecular pathway and transduce signals through the ERM-dependent machinery that is responsible for the remodeling and stabilization of the synapse. Cdc42 activity is impaired in β1 integrin-deficient T cells that form conjugates with antigen-presenting cells but is partially restored in the context of an antigen-specific synapse. This restoration of Cdc42 activity is due, at least in part, to the recruitment and activation of β2 integrin.

Adaptive fields: distributed representations of classically conditioned associations

Present neural models of classical conditioning all suffer from the same shortcoming: local representation of information (therefore, very precise neural prewiring is necessary). As an alternative the authors develop two neural models of classical conditioning which rely on distributed representations of information. Both models are of the Hopfield type. In the first model the existence of transmission delays is used to store temporal relations. The second model is based on interactions between spatially separated neural fields. Using tools from statistical mechanics the authors show that behavioural constraints can be met only if the Hebb rule is extended with inter- or intrasynaptic competition.

Finite connectivity attractor neural networks

We study a family of diluted attractor neural networks with a finite average number of (symmetric) connections per neuron. As in finite connectivity spin glasses, their equilibrium properties are described by order parameter functions, for which we derive an integral equation in replica symmetric approximation. A bifurcation analysis of this equation reveals the locations of the paramagnetic to recall and paramagnetic to spin-glass transition lines in the phase diagram. The line separating the retrieval phase from the spin-glass phase is calculated at zero temperature. All phase transitions are found to be continuous.

Bayesian analysis of fluorescence lifetime imaging data

Fluorescence Lifetime Imaging (FLIM) is an intensity independent and sensitive optical technique for studying the cellular environment but its accuracy is often compromised when low photon counts are available for analysis. We have developed a photon-by-photon Bayesian analysis method targeted at the accurate analysis of low photon count time-domain FLIM data collected using Time Correlated Single Photon Counting (TCSPC). Parameter estimates obtained with our mono-exponential Bayesian analysis compare favorably with those using maximum likelihood, least squares, and phasor analysis, offering robust estimation with greater precision at very low total photon counts, particularly in the presence of significant background levels. Details of the Bayesian implementation are presented alongside results of mono-exponential analysis of both real and synthetic data. We demonstrate that for low photon count data, obtained by imaging human epithelial carcinoma cells expressing cdc42-GFP, Bayesian analysis estimates the green fluorescent protein (GFP) lifetime to a level of accuracy not obtained using maximum likelihood estimation or other techniques. These results are echoed by the analysis of synthetic decay data incorporating a 10% uniform background, with our Bayesian analysis routines yielding lifetime estimates within an accuracy of 20% with about 50 counts. This level of precision is not achieved with maximum likelihood nor phasor analysis techniques with fewer than 100 counts.

A Bayesian method for single molecule, fluorescence burst analysis

There is currently great interest in determining physical parameters, e.g. fluorescence lifetime, of individual molecules that inform on environmental conditions, whilst avoiding the artefacts of ensemble averaging. Protein interactions, molecular dynamics and sub-species can all be studied. In a burst integrated fluorescence lifetime (BIFL) experiment, identification of fluorescent bursts from single molecules above background detection is a problem. This paper presents a Bayesian method for burst identification based on model selection and demonstrates the detection of bursts consisting of 10% signal amplitude. The method also estimates the fluorescence lifetime (and its error) from the burst data.

Coupled dynamics of fast spins and slow interactions in neural networks and spin systems

The authors examine an Ising spin system in which both spins and the interactions between them may evolve in time, although on disparate timescales, such that the couplings change adiabatically. In thermal equilibrium they find a novel application of the replica method, but for finite replica number, representing the ratio of the temperatures of the spin and interaction systems. Regimes where the motion of the couplings has non-trivial effects are found in addition to those where solely the stochasticity of these interaction weights in significant, and this issue is closely related to the orders of the transitions between the various phases observed. Simulation results lend support to the analysis.

Delays in Neural Networks

In this letter we consider the effect of random transmission delays on the dynamics of a fully connected neural network. We show that, if these delays are also present during a learning stage in which patterns are presented in succession, the network will be capable of regenerating this sequence of patterns. This capability does not depend on the actual distribution of the delays. For the condition where the pattern sequence has a nonzero correlation time we present an equation with which to compute the time-dependent overlap of the system state with the sequence.