Biology subjects

Lin, M. M.

Publications and source records attributed to Lin, M. M..

14 recordsLinked to original sources

bioRxiv2026

Hidden Markov Modeling Reveals Altered Functional Network Dynamics in Multiple Sclerosis

Purpose: Multiple sclerosis (MS) is a neurodegenerative and demyelinating disease of the central nervous system. Recent evidence suggests that localized gray matter atrophy in MS may be involved in distributed functional brain networks. However, the dynamic trajectory of functional network alterations in MS remains unclear. The current study extends prior network-mediated findings and aims to investigate the dynamic functional connectivity (dFC) and energetic landscape across an atrophy-prone network in MS. Methods: Twenty MS participants and twenty healthy controls (HC) underwent resting-state fMRI. The atrophy-based functional network (AFN) model was implemented, and time series from atrophy-prone regions were fit to a 12-state hidden Markov model. Each functional network state, a recurring pattern of activity across the AFN, was characterized by the AFN regions with the strongest activation. The states were then grouped into two metastates. Dynamic functional network metrics were compared between groups by permutation testing and correlated with disability and cognition. The energetic landscape of the network was mapped with a pairwise maximum-entropy model. Temporal irreversibility was also measured, which defines the arrow-of-time asymmetry of network dynamics. Results: People with MS showed lower occupancy of state 1, which peaked on the right middle frontal and anterior cingulate cortex. Occupancy and mean lifetime were higher in states 3 and 10 (p < 0.05). In state 3, cortical activity was below the network average, and state 10 peaked on the postcentral gyrus. At the metastate level, MS participants spent more time in metastate M1 and less in M2 (p = 0.04). Within the MS group, a longer interval between visits to state 3 was associated with greater disability (r = 0.57), and a higher switching rate in state 10 with poorer verbal fluency (r = -0.50). The energetic landscape was similar between groups. Temporal irreversibility was lower in MS, mainly across subcortical connections. Conclusion: MS participants dwelled longer in specific dFC states of the AFN, while the rate of transition between states did not differ from healthy controls. Dynamic functional network metrics were associated with disability and cognitive performance and may serve as functional correlates of clinical impairment. The states were similar between groups, whereas the directed drive between states was lower across subcortical connections. This pattern suggests altered network dynamics rather than altered network architecture in MS. Findings from this study characterize the functional network dynamics of neurodegeneration in MS, which may inform future development of network-based approaches for monitoring disease progression.

neuroscience↗

bioRxiv2025

Saturation mutagenesis map of generalist versus specialist adaptations of β-lactamase to novel antibiotics

The evolution of {beta}-lactamase proteins is shaped by the need to maintain enzymatic activity against previously prevalent {beta}-lactam antibiotics while expanding substrate range against new classes of antibiotics. Using saturation mutagenesis and sequence-barcoding-based quantification, we comprehensively mapped the response of the fitness landscape of TEM-1 {beta}-lactamase, which evolved against penicillin-class antibiotics, to mutational perturbations against six diverse {beta}-lactam antibiotics. This systematic panel of antibiotic substrates, including representatives from penicillin, cephalosporin, and monobactam classes, allowed us to classify resistance mutations into two categories. Generalist mutations conferred resistance to multiple antibiotics and were consistently restricted to three positions critical for substrate recognition and catalytic function R164, G238, and E240. These substitutions produced broad spectrum resistance through mechanisms such as expansion of the active site and improved substrate accommodation. In contrast, specialist mutations conferred resistance to only a single antibiotic and exhibited much wider positional diversity. Ceftazidime selection yielded the greatest number of distinct specialist mutations, which were frequently found in flexible or peripheral regions including the omega loop. One especially unexpected finding was the identification of the E166P variant. E166 is a catalytic residue required for deacylation during hydrolysis, and substitutions at this site are generally assumed to abolish function. However, E166P conferred a significant increase in ceftazidime resistance despite eliminating activity against penicillins. Molecular dynamics simulations and mutational analysis revealed that the E166P mutant employs an alternative catalytic mechanism, involving residue N132, rather than the canonical pathway. Together, our findings reveal, at the molecular level, how specialist mutations open up a wide range of diverse and idiosyncratic solutions at the expense of generalizability. These insights may inform strategic design of antibiotic administration protocols to systematically lower pathogenic evolvability.

evolutionary biology↗

bioRxiv2025

Discrete protein dynamics enable long-range communication

Allostery, in which perturbations at an input protein site tune the activity at a distant output site, allows proteins to serve as molecular logic gates. Often, information is transmitted without altering the structure outside of the input and output sites. This focalized allostery requires correlated motion between protein backbone dihedral angles that are separated by distances many times longer than the scale of electrostatic interactions. What physical properties of folded proteins enable such long-distance information sharing despite thermal noise is unclear. To address this question, we introduce a Variable-Well Dihedral (VWD) model Hamiltonian which removes dependence on chemical details and forces, instead only tuning the degree of nonlinearity of purely local interactions within a densely-packed polymer. We show that tuning the physical parameters of the model gives rise to focalized allostery in so far that doing so increases the discreteness of the internal degrees of freedom, with real proteins occupying the highly discrete regime. These results parallel, at the molecular scale, the superiority of digital compared to analog signal processing for telecommuncations under noisy conditions.

biophysics↗

bioRxiv2025

Mechanism of MEK1 activation by phosphorylation

Phosphorylation is the most common post-translational protein modification, often acting as the activator of protein function. In the case of MEK1, a member of the MAP kinase family, phosphorylation of both S218 and S222 are necessary for activation. Yet the molecular activation mechanism and its cooperative nature are poorly understood, especially due to the lack of experimental phosphorylated MEK1 structures. We performed molecular dynamics simulations to investigate the structural and dynamical consequences of single and double phosphorylation in MEK1. We find that successive phosphorylation progressively unwinds the helix containing the phosphorylation sites, thereby rotating the phosphate groups to directly interact with the catalytic site. Consequently, the solvent accessible surface area of the catalytic residues increase with phosphorylation. Yet, only in the double-phosphorylated state do all four critical catalytic residues become solvent exposed. By calculating the conformational entropy, we find that only in the double-phosphorylated state do all four catalytic residues have mutual information with each other, suggesting that phosphorylation also induces correlated motion at the active site. To validate our approach, we show that our simulations predict the alignment of the R-spine motif in the double-phosphorylated state, in agreement with the other kinases for which this alignment has been experimentally observed. These results show that phosphorylation can, via a partial unfolding mechanism, increase the solvent exposure of and dynamical coupling within the active site, both of which are critical for enzymatic activity.

biophysics↗

bioRxiv2025

Higher-order epistasis drives evolutionary unpredictability toward novel antibiotic resistance

The rapid evolution of extended-spectrum {beta}-lactamases (ESBLs) represents a global health threat, undermining the efficacy of {beta}-lactams, the most extensively used antibiotic class. To elucidate the evolutionary dynamics underlying {beta}-lactam resistance, we constructed a comprehensive combinatorial mutant library comprising all 55,296 possible TEM-1 {beta}-lactamase variants integrating 18 clinically observed mutations across 13 key residues. Over eight million empirical fitness measurements were obtained under selection pressure with both a native antibiotic substrate (ampicillin) and a novel antibiotic (aztreonam), establishing the largest experimentally determined fitness landscape for antibiotic resistance to date. Through graph-theoretic and epistatic analyses, we discovered that selection with ampicillin resulted in weak epistasis, with mutants rarely surpassing the fitness of the wild-type enzyme. Conversely, aztreonam selection elicited extensive higher-order epistasis, generating a rugged fitness landscape characterized by increased phenotypic unpredictability. Interpretable machine-learning analyses identified context-dependent epistatic interactions necessary for achieving high-level aztreonam resistance. Further evolutionary statistical analyses, including direct coupling analysis and latent generative landscapes, showed that top-performing TEM-1 variants consistently adhered to conserved epistatic patterns found in naturally occurring {beta}-lactamases. Our findings demonstrate that higher-order epistasis critically shapes fitness landscape ruggedness when enzymes adapt to novel substrates, whereas adaptations to native substrates exhibit predictably smoother landscapes. This integrated experimental and computational framework provides a foundation for predictive evolutionary pharmacology, enabling assessments of newly developed {beta}-lactams or emerging {beta}-lactamase variants for their potential contribution to ESBL evolution. Importantly, incorporating graph-theoretically informed evolutionary constraints can strategically disrupt evolutionary pathways, presenting a viable approach to mitigate the rise of antibiotic resistance.

evolutionary biology↗

bioRxiv2025

Entropy tree networks of residue dynamics encode protein allostery

To respond to environmental cues, proteins must amplify angstrom-scale signals across nanometers in the presence of thermal fluctuations. A prevailing view is that thermal fluctuations attenuate (1-7) signal-bearing coherent motions (8-11), yet numerous experiments correlate signaling state with fluctuations themselves (12-36). Here, we show that residue-level fluctuations encode "geometric bits" that are communicated within a sparse 3D network of shared entropy. We demonstrate this by developing an open-source framework that discovers shared entropy networks by inferring discrete residue conformations from molecular dynamics simulations of protein structure and finding maximum-likelihood tree distributions with minimal assumptions, enabling multiscale conformational entropy calculation without exhaustive enumeration. We validate our approach against an array of experimental data modalities probing sequence and ligand-dependent functions of PDZ and estrogen receptor ligand-binding domains, accurately predicting allosteric hotspots in saturation mutagenesis with residue-scale resolution, local entropies correlated with NMR and HDX dynamics, and global entropies in agreement with calorimetry without fitting. Comparing networks of the six human steroid receptors recovers phylogenetic history, providing evidence that evolution achieves functional diversity by reprogramming entropy within a fixed protein fold. The ability to transmit signals by harnessing thermal fluctuations categorically distinguishes proteins from human-designed communication channels, for which fluctuations are noise to be minimized.

biophysics↗

bioRxiv2025

An invariant schema emerges within a neural network during hierarchical learning of visual boundaries

Neural circuits must balance plasticity and stability to enable continual learning without catastrophic forgetting, a pervasive feature of artificial neural networks trained using end-to-end learning (e.g. backpropagation). Here, we apply an alternative, hierarchical learning algorithm to the cognitive task of boundary detection in video clips. In contrast to backpropagation, hierarchical training converges to a network executing a fixed schema and generates firing statistics consistent with single-neuron recordings from human subjects performing the same task. The hierarchically trained networks schema circuit remains invariant following training on sparse data, with additional data serving to refine the upstream representation.

neuroscience↗

bioRxiv2024

Enhanced Pooling for Weakly Supervised Gigapixel WSI Training Improves Classification and Lesion Localization

A robust artificial intelligence-assisted workflow for tumor assessment in pathology requires not only accurate classification but also precise lesion localization. While current weakly supervised learning methods significantly reduce the need for extensive annotations and leverage large quantities of annotation-free whole-slide images (WSIs) to enhance classification robustness, they often fall short in segmentation accuracy. We attribute this limitation to the optimization goals in classification, which tend to focus solely on the most representative features--an approach that is particularly inefficient for WSIs with gigapixel resolution. To address this challenge, we introduce a novel approach based on streaming convolution, an end-to-end method for WSI training. Our contributions include the Rectified LogSumExp (ReLSE) pooling method and adaptive pseudo annotation generation for self-training, both designed to encourage models to learn from sub-representative features. Using only slide-level annotations from the CAMELYON16 dataset, our method achieves a significant improvement in metastasis localization, with a patch-level recall from 53.88% to 79.18% at a precision of 80%. This conclusion also holds for a dataset collected from Taipei Veterans General Hospital (TVGH), used in the assessment of lung cancer lymph node metastasis with a recall improved from 42.98% to 72.27%. The proposed method can also be integrated with additional manual annotations. Experiment results showed the semi-supervised model (recall: 91.12%) could even outperform the strongly supervised model (recall: 81.80%) on the TVGH dataset.

bioinformatics↗

bioRxiv2024

A lever hypothesis for Synaptotagmin-1 action in neurotransmitter release

Neurotransmiter release is triggered in microseconds by Ca2+-binding to the Synaptotagmin-1 C2 domains and by SNARE complexes that form four-helix bundles between synaptic vesicles and plasma membranes, but the coupling mechanism between Ca2+-sensing and membrane fusion is unknown. Release requires extension of SNARE helices into juxtamembrane linkers that precede transmembrane regions (linker zippering) and binding of the Synaptotagmin-1 C2B domain to SNARE complexes through a primary interface comprising two regions (I and II). The Synaptotagmin-1 Ca2+-binding loops were believed to accelerate membrane fusion by inducing membrane curvature, perturbing lipid bilayers or helping bridge the membranes, but SNARE complex binding orients the Ca2+-binding loops away from the fusion site, hindering these putative activities. Molecular dynamics simulations now suggest that Synaptotagmin-1 C2 domains near the site of fusion hinder SNARE action, providing an explanation for this paradox and arguing against previous models of Sytnaptotagmin-1 action. NMR experiments reveal that binding of C2B domain arginines to SNARE acidic residues at region II remains after disruption of region I. These results and fluorescence resonance energy transfer assays, together with previous data, suggest that Ca2+ causes reorientation of the C2B domain on the membrane and dissociation from the SNAREs at region I but not region II. Based on these results and molecular modeling, we propose that Synaptotagmin-1 acts as a lever that pulls the SNARE complex when Ca2+ causes reorientation of the C2B domain, facilitating linker zippering and fast membrane fusion. This hypothesis is supported by the electrophysiological data described in the accompanying paper. Significance statementNeurotransmiter release requires SNARE complexes that fuse synaptic vesicles with the plasma membrane and the Ca2+-sensor synaptotagmin-1, which was thought to facilitate membrane fusion directly through its Ca2+-binding loops. However, binding of Synaptotagmin-1 to SNARE complexes orients these loops away from the fusion site. Using molecular dynamics simulations, we show that placing Synaptotagmin-1 at the fusion site hinders the action of SNARE complexes. Spectroscopic studies show that Ca2+ binding to Synaptotagmin-1 can change its interactions with SNARE complexes and, together with molecular modeling, suggest that Synaptotagmin-1 acts as a lever, pulling SNARE complexes and thus facilitating their action on the membranes to induce fusion. Functional studies described in the accompanying paper support this hypothesis.

biophysics↗

bioRxiv2023

Molecular mechanism underlying SNARE-mediated membrane fusion enlightened by all-atom molecular dynamics simulations

The SNARE proteins syntaxin-1, SNAP-25 and synaptobrevin mediate neurotransmiter release by forming tight SNARE complexes that fuse synaptic vesicles with the plasma membranes in microseconds. Membrane fusion is generally explained by the action of proteins on overall membrane properties such as curvature, elastic modulus and tension, and a widespread model envisions that the SNARE motifs, juxtamembrane linkers and C-terminal transmembrane regions of synaptobrevin and syntaxin-1 form continuous helices that act mechanically as semi-rigid rods, squeezing the membranes together as they assemble ( zipper) from the N- to the C-termini. However, the mechanism underlying fast SNARE-induced membrane fusion remains unknown. We have used all-atom molecular dynamics simulations to investigate this mechanism. Our results need to be interpreted with caution because of the limited number and length of the simulations, but they suggest a model of membrane fusion that has a natural physicochemical basis, emphasizes local molecular events over general membrane properties, and explains extensive experimental data. In this model, the central event that initiates fast (microsecond scale) membrane fusion occurs when the SNARE helices zipper into the juxtamembrane linkers which, together with the adjacent transmembrane regions, promote encounters of acyl chains from both bilayers at the polar interface. The resulting hydrophobic nucleus rapidly expands into stalk-like structures that gradually progress to form a fusion pore, aided by the SNARE transmembrane regions and without clearly discernible intermediates. The propensity of polyunsaturated lipids to participate in encounters that initiate fusion suggests that these lipids may be important for the high speed of neurotransmiter release.

biophysics↗

bioRxiv2022

Kinetic and thermodynamic allostery in the Ras protein family

Allostery, the tranfer of information between distant parts of a macromolecule, is a fundamental feature of protein function and regulation. However, allosteric mechanisms are usually not explained by protein structure, requiring information on correlated fluctuations uniquely accessible to molecular simulation. Existing work to extract allosteric pathways from molecular dynamics simulations has focused on thermodynamic correlations. Here we show how kinetic correlations (i.e. dynamical activity) encode complementary information essential to explain observed variations in allosteric regulation. We performed atomistic simulations, totalling 0.5 milliseconds, on H, K, and NRas isoforms in the apo, GTP, and GDP-bound states of Ras protein, with and without complexing to its downstream effector, Raf. We show that differences in experimentally measured intrinsic and Raf-dependent catalytic speed amongst the three isoforms can be explained by dynamical activity and entropy, respectively. We show that Switch I and Switch II are the primary components of thermodynamic and kinetic allosteric networks, consistent with the key roles of these two motifs. These communication networks are altered by the hydrolysis of the substrate gamma phosphate, leading to increased entropy in HRas loops involved in substrate release. We find that the putative allosteric region is not coupled in KRas, but is coupled to the hydrolysis arm switch II in NRas and HRas, and that the mechanism in the latter two isoforms are thermodynamic and kinetic, respectively. Binding of Raf-RBD further activates thermodynamic allostery in HRas and KRas but has limited effect on NRas. These results indicate that kinetic and thermodynamic correlations are both needed to explain protein function and allostery. These two distinct channels of allosteric regulation, and their combinatorial variability, may explain how subtle mutational differences can lead to diverse regulatory profiles among enzymatic proteins.

biophysics↗

bioRxiv2022

Resonant fluctuations of selection pressure exponentially accelerate fitness valley crossing

Two functional protein sequences can sometimes be separated by a fitness valley - a series of low or non-functional intermediate mutations that must be traversed to reach a more optimal or refined function. Time-varying selection pressure modulates evolutionary sampling of such valleys. Yet, how the amplitude and frequency of fluctuating selection influence the rate of protein evolution is poorly understood. Here, we derive a simple equation for the time-dependent probability of crossing a fitness valley as a function of evolutionary parameters: valley width, protein size, mutation rate, and selection pressure. The equation predicts that, under low selection pressure, the valley crossing rate is magnified by a factor that depends exponentially on valley width. However, after a characteristic time set by the evolutionary parameters, the rate rapidly decays. Thus, there is an optimal frequency of selection-pressure fluctuations that maximizes the rate of protein optimization. This result is reminiscent of the resonance frequency in mechanical systems. The equation unites empirical and theoretical results that were previously disconnected, and is consistent with time-dependent in vitro and clinical data. More generally, these results suggest that seasonal and climate oscillations could synchronously drive protein evolution at the resonant frequency across a range of organism hosts and timescales. This theory could also be applied to optimize de novo protein evolution in laboratory directed evolution using time-varying protocols.

evolutionary biology↗

bioRxiv2021

All-atom molecular dynamics simulations of synaptic vesicle fusion I: a glimpse at the primed state

Synaptic vesicles are primed into a state that is ready for fast neurotransmitter release upon Ca2+-binding to synaptotagmin-1. This state likely includes trans-SNARE complexes between the vesicle and plasma membranes that are bound to synaptotagmin-1 and complexins. However, the nature of this state and the steps leading to membrane fusion are unclear, in part because of the difficulty of studying this dynamic process experimentally. To shed light into these questions, we performed all-atom molecular dynamics simulations of systems containing trans-SNARE complexes between two flat bilayers or a vesicle and a flat bilayer with or without fragments of synaptotagmin-1 and/or complexin-1. Our results help visualize potential states of the release machinery en route to fusion, and suggest mechanistic features that may control the speed of release. In particular, the simulations suggest that the primed state contains almost fully assembled trans-SNARE complexes bound to the synaptotagmin-1 C2B domain and complexin-1 in a spring-loaded configuration where interactions of the C2B domain with the plasma membrane orient complexin-1 toward the vesicle, avoiding premature membrane merger but keeping the system ready for fast fusion upon Ca2+ influx.

biophysics↗

bioRxiv2020

All protein aggregates will thermodynamically order

Proteins can aggregate into disordered liquid droplets or ordered assemblies such as amyloid fibrils. These two distinct phases determine the spatial organization within cells and serve differing roles in a wide range of biological functions including gene regulation, organelle and synapse formation, and memory consolidation. The ordered phase can also give rise to diseases such as Alzheimers. However, how the protein sequence determines aggregation fate is an open question. Here we establish a general statistical mechanical theory of the disordered-to-ordered transition for polymer aggregates, including proteins, thereby completing the phase diagram for this general class of matter. The theory produces a simple universal equation determining the favored phase as a function of the temperature, polymer length, and inter-residue interaction energy variance. We show that the sequence-dependent energy variance can be efficiently calculated from all-atom molecular dynamics simulations, so that the theory has no adjustable parameters. The equation accurately predicts the experimental length-dependent crystallization temperature of synthetic polymers. The theory shows that all protein aggregates, regardless of sequence, will thermodynamically order, even in the most extreme thermophiles. Therefore, energy must be expended to maintain the disordered protein aggregate at steady state. More broadly, the theory establishes a lower bound on the ordering transition temperature for polymers. This bound indicates that condensates of any organic polymer will spontaneously order on all habitable planets.

biophysics↗