Publications

TSC6-peptide is an 8-amino acid synthetic peptide compound that inhibits intestinal α-amylase upon oral administration to control the digestion and absorption of starchy foods. However, its toxicological safety and other scientific evidence remain to be established. Given the potential utility of TSC6-peptide as an α-amylase inhibitor, this study conducted in vivo and in vitro experiments to support its safety profile. In reverse mutation and mammalian micronucleus assays revealed no genotoxicity for TSC6-peptide. Meanwhile, in the 90-day dietary toxicity study in rats, the NOAEL was 500 mg/kg bw/day for male rats following TSC6-peptide administration and the LOAEL for female rats was 50 mg/kg bw/day. These results indicate that TSC6-peptide possesses controllable safety as an α-amylase inhibitor.

Target-conditioned de novo siRNA design requires capturing both target mRNA binding context and internal siRNA sequence-activity rules. Traditional computational pipelines predominantly rely on discriminative prediction models that rank pre-filtered candidate pools. However, these approaches struggle with generalization on novel target genes and often overlook potent candidates due to dataset size constraints and motif overfitting. To reconcile generation precision with sequence diversity, we propose siDiff, an efficacy-guided discrete diffusion framework for target-conditioned de novo siRNA design. siDiff pairs a discrete-masked diffusion transformer—which models the underlying sequence distribution over functional duplexes—with a mask-robust efficacy guidance model. During inference, we introduce a biology-aware, three-stage sampling mechanism that performs structural candidate filtering, dynamic unmasking guidance, and cluster-aware redundancy mitigation. This dual mechanism enables the diffusion process to explore broad sequence spaces while the efficacy model prevents distributional shift toward non-functional candidates. Extensive experiments across four datasets, including the public Takayuki benchmark and three curated patent datasets, demonstrate that siDiff significantly outperforms state-of-the-art discriminative baselines and discrete diffusion models, achieving relative hit-rate improvements of over 30% and demonstrating superior generalization on out-of-distribution gene targets. The source code and related materials are available at https://github.com/cybericha/siDiff.

Predicting single-cell transcriptional responses to genetic, chemical and cytokine perturbations is a fundamental challenge in computational biology and AI Virtual Cell (AIVC) modeling, with direct implications for drug discovery and the elucidation of gene regulatory networks. Existing approaches often rely on auxiliary cell-state encoders, hierarchical variational autoencoders, dedicated Transformer encoder-decoder modules, or gene-interaction priors to compress high-dimensional expression profiles into latent representations. While effective, these designs increase architectural complexity and may limit scalability and generalizability. This paper introduces OCOO-T 1, a minimalist flow-matching-based AIVC model for transcriptional perturbation response prediction. OCOO-T utilizes a vanilla Transformer stack that operates directly on continuous gene expression profiles and formulates perturbation response prediction as a continuous-time denoising process. Perturbation embeddings, dosage information, and cell-line/cell-type specificity are integrated through adaptive layer normalization and in-context tokens. Comprehensive evaluations on Tahoe100M, Replogle, and PBMC benchmarks demonstrate that OCOO-T achieves state-of-the-art performance across diverse perturbations and cell types while effectively scaling to long transcriptional profiles through patching and depatching of cellular contexts. By leveraging the simplicity of Transformer-based denoising for single-cell omics, OCOO-T provides an effective and scalable framework for in-silico cellular simulation.

Small interfering RNA (siRNA) is a short double-stranded RNA molecule (about 21-23 nucleotides) with the potential to cure diseases by silencing the function of target genes. Due to its well-understood mechanism, many siRNA-based drugs have been evaluated in clinical trials. However, selecting effective binding regions and designing siRNA sequences requires extensive experimentation, making the process costly. As genomic resources and publicly available siRNA datasets continue to grow, data-driven models can be leveraged to better understand siRNA-mRNA interactions. To fully exploit such data, curating high-quality siRNA datasets is essential to minimize experimental errors and noise. We propose siDPT: siRNA efficacy Prediction via Debiased Preference-Pair Transformer, a framework that constructs a preference-pair dataset and designs an siRNA-mRNA interactive transformer with debiased ranking objectives to improve siRNA inhibition prediction and generalization. We evaluate our approach using two public datasets and one newly collected patent dataset. Our model demonstrates substantial improvement in Pearson correlation and strong performance across other metrics.

Messenger RNA (mRNA)-based vaccines are accelerating the discovery of new drugs and revolutionizing the pharmaceutical industry. However, selecting particular mRNA sequences for vaccines and therapeutics from extensive mRNA libraries is costly. Effective mRNA therapeutics require carefully designed sequences with optimized expression levels and stability. This paper proposes a novel contextual language model (LM)-based embedding method: mRNA2vec. In contrast to existing mRNA embedding approaches, our method is based on the self-supervised teacher-student learning framework of data2vec. We jointly use the 5' untranslated region (UTR) and coding sequence (CDS) region as the input sequences. We adapt our LM-based approach specifically to mRNA by 1) considering the importance of location on the mRNA sequence with probabilistic masking, 2) using Minimum Free Energy (MFE) prediction and Secondary Structure (SS) classification as additional pretext tasks. mRNA2vec demonstrates significant improvements in translation efficiency (TE) and expression level (EL) prediction tasks in UTR compared to SOTA methods such as UTR-LM. It also gives a competitive performance in mRNA stability and protein production level tasks in CDS such as CodonBERT.

Drug discovery is protracted, resource-intensive, and afflicted by attrition rates exceeding 90 %, which leaves most diseases, particularly rare or neglected indications, without effective therapies. Drug repurposing offers a cost effective alternative, yet systematic identification of novel drug indication pairs and mechanistic rationales remains hindered by the scale and heterogeneity of biomedical knowledge. We present BioScientist Agent, an end to end framework that unifies a billion-fact biomedical knowledge graph with (i) a variational graph auto-encoder for representation learning and link prediction driven drug repurposing, (ii) a reinforcement learning module that traverses the graph to recover biologically plausible mechanistic paths, and (iii) A large language model (LLM) multi-agent layer that orchestrates these components, enabling inference of target pathways for a drug disease pair, and automatic generation of coherent causal reports. In all downstream tasks, the BioScientist Agent surpasses existing state of the art baseline models across various metrics and provides mechanistic explanations consistent with the literature. Its open and modular design accelerates hypothesis generation and reduces experimental overhead in early stage discovery.

High-content analysis (HCA) holds enormous potential for drug discovery and research, but widely used methods can be cumbersome and yield inaccurate results. Noisy and redundant signals in cell images impede accurate deep learning-based image analysis. To address these issues, we introduce X-Profiler, a novel HCA method that combines cellular experiments, image processing, and deep learning modeling. X-Profiler combines the convolutional neural network and Transformer to encode high-content images, effectively filtering out noisy signals and precisely characterizing cell phenotypes. In comparative tests on drug-induced cardiotoxicity, mitochondrial toxicity classification, and compound classification, X-Profiler outperformed both DeepProfiler and CellProfiler, as two highly recognized and representative methods in this field. Our results demonstrate the utility and versatility of X-Profiler, and we anticipate its wide application in HCA for advancing drug development and disease research.

B-cell epitope identification plays a vital role in the development of vaccines, therapies, and diagnostic tools. Currently, molecular docking tools in B-cell epitope prediction are heavily influenced by empirical parameters and require significant computational resources, rendering a great challenge to meet large-scale prediction demands. When predicting epitopes from antigen-antibody complex, current artificial intelligence algorithms cannot accurately implement the prediction due to insufficient protein feature representations, indicating novel algorithm is desperately needed for efficient protein information extraction. In this paper, we introduce a multimodal model called WUREN (Whole-modal Union Representation for Epitope predictioN), which effectively combines sequence, graph, and structural features. It achieved AUC-PR scores of 0.213 and 0.193 on the solved structures and AlphaFold-generated structures, respectively, for the independent test proteins selected from DiscoTope3 benchmark. Our findings indicate that WUREN is an efficient feature extraction model for protein complexes, with the generalizable application potential in the development of protein-based drugs. Moreover, the streamlined framework of WUREN could be readily extended to model similar biomolecules, such as nucleic acids, carbohydrates, and lipids.

AO_SCPLOWBSTRACTC_SCPLOWThe rapid advancement of single-cell sequencing technology has significantly deepened our understanding of cellular heterogeneity, yet it concurrently presents substantial challenges for the unified modeling of single-cell data. Simultaneously, pre-trained foundation models have achieved notable success in domains such as natural language processing and image analysis. However, extending these models to accommodate ultra-long single-cell transcriptome sequences, characterized by an extensive number of genes, remains a formidable task. In this study, we introduce SC-MAMBA2, based on the MAMBA2 architecture, meticulously designed with a bidirectional modeling approach tailored for single-cell transcriptomics data. As the first single-cell foundation model to integrate state-space models (SSMs) underlying MAMBA2 architecture, SC-MAMBA2 features over 625 million parameters, covers more than 60,000 genes, and was pre-trained on a dataset of over 57 million cells, making it the most comprehensive solution for processing ultra-long transcriptome sequences. Extensive bench-marking across a diverse array of downstream tasks consistently demonstrates that SC-MAMBA2 surpasses state-of-the-art models, delivering superior accuracy and enhanced computational efficiency.

Norepinephrine transporter (NET; encoded by SLC6A2) reuptakes the majority of the released noradrenaline back to the presynaptic terminals, thereby affecting the synaptic noradrenaline level1. Genetic mutations and dysregulation of NET are associated with a spectrum of neurological conditions in humans, making NET an important therapeutic target1. However, the structure and mechanism of NET remain unclear. Here we provide cryogenic electron microscopy structures of the human NET (hNET) in three functional states-the apo state, and in states bound to the substrate meta-iodobenzylguanidine (MIBG) or the orthosteric inhibitor radafaxine. These structures were captured in an inward-facing conformation, with a tightly sealed extracellular gate and an open intracellular gate. The substrate MIBG binds at the centre of hNET. Radafaxine also occupies the substrate-binding site and might block the structural transition of hNET for inhibition. These structures provide insights into the mechanism of substrate recognition and orthosteric inhibition of hNET.

Design of molecules for candidate compound selection is one of the central challenges in drug discovery due to the complexity of chemical space and requirement of multi-parameter optimization. Here we present an application scenario-oriented platform (ID4Idea) for molecule generation in different scenarios of drug discovery. This platform utilizes both library or rule based and generative based algorithms (VAE, RNN, GAN, etc.), in combination with various AI learning types (pre-training, transfer learning, reinforcement learning, active learning, etc.) and input representations (1D SMILES, 2D graph, 3D shape, binding site, pharmacophore, etc.), to enable customized solutions for a given molecular design scenario. Besides the usual generation followed screening protocol, goal-directed molecule generation can also be conducted towards predefined goals, enhancing the efficiency of hit identification, lead finding, and lead optimization. We demonstrate the effectiveness of ID4Idea platform through case studies, showcasing customized solutions for different design tasks using various input information, such as binding pockets, pharmacophores, and compound representations. In addition, remaining challenges are discussed to unlock the full potential of AI models in drug discovery and pave the way for the development of novel therapeutics.

Purpose: Recently, there has been rapid development in model-informed drug development, which has the potential to reduce animal experiments and accelerate drug discovery. Physiologically based pharmacokinetic (PBPK) and machine learning (ML) models are commonly used in early drug discovery to predict drug properties. However, basic PBPK models require a large number of molecule-specific inputs from in vitro experiments, which hinders the efficiency and accuracy of these models. To address this issue, this paper introduces a new computational platform that combines ML and PBPK models. The platform predicts molecule PK profiles with high accuracy and without the need for experimental data. Methods: This study developed a whole-body PBPK model and ML models of plasma protein fraction unbound (f up), Caco-2 cell permeability, and total plasma clearance to predict the PK of small molecules after intravenous administration. Pharmacokinetic profiles were simulated using a "bottom-up" PBPK modeling approach with ML inputs. Additionally, 40 compounds were used to evaluate the platform's accuracy. Results: Results showed that the ML-PBPK model predicted the area under the concentration-time curve (AUC) with 65.0 % accuracy within a 2-fold range, which was higher than using in vitro inputs with 47.5 % accuracy. Conclusion: The ML-PBPK model platform provides high accuracy in prediction and reduces the number of experiments and time required compared to traditional PBPK approaches. The platform successfully predicts human PK parameters without in vitro and in vivo experiments and can potentially guide early drug discovery and development.

In the realm of drug discovery, the Design-Make-Test-Analyses (DMTA) cycle plays a pivotal role, with the “M” phase often emerging as a bottleneck. Numerous endeavors have sought to expedite this phase, employing technologies like robotic arms, flow chemistry, and other devices for automated synthesis. This paper introduces “preMTA” as a game-changer, leveraging artificial intelligence (AI)-based molecular generative models, automated synthesis workflows, and the ASMS platform, creating the “D-preMTA-MTA” cycle. During the “preMTA” phase, AI-designed chemical spaces for specific targets are explored. Microgram-scaled target compounds are efficiently prepared through pooling reactions and streamlined work-ups via automated synthesis workflows. These compounds are then screened using ASMS to rank them based on competitive binding affinity. This strategic approach enables the subsequent MTA phase to focus on potent binders. Validation of this approach involved the discovery of novel inhibitors for T-cell protein tyrosine phosphatase (TCPTP or PTPN2). In the “D” phase, a target-focused library of 696 compounds was designed, with 140 strong binders swiftly identified in the “preMTA” phase. Subsequently, 51 of these compounds were scale-up synthesized and confirmed with IC 50 values ranging from 16 nM to 277 nM, with 17 exhibiting IC 50 values below 50 nM. The efficacy and seamless integration of each phase in the “D-preMTA-MTA” cycle can be attributed to a) the use of a synthesis-oriented molecular generative method streamlining automated synthesis, b) the deployment of a versatile and adaptable robotic arm capable of multitasking, and c) the implementation of ASMS screening methods reducing both synthetic and bio-testing complexities.

Introduction: Antibodies represent a specific class of proteins produced by the adaptive immune system in response to pathogens. Mining the information embedded in antibody amino acid sequences can benefit both antibody property prediction and novel therapeutic development. However, antibodies possess unique features that should be incorporated using specifically designed training methods, leaving room for improvement in pre-training models for antibody sequences. Methods: In this study, we present a Pre-trained model of Antibody sequences trained with a Rational Approach for antibodies (PARA). PARA employs a strategy conforming to antibody sequence patterns and an advanced natural language processing self-encoding model structure. This approach addresses the limitations of existing protein pre-training models, which primarily utilize language models without fully considering the differences between protein sequences and language sequences. Results: We demonstrate PARA's performance on several tasks by comparing it to various published pre-training models of antibodies. The results show that PARA significantly outperforms existing models on these tasks, suggesting that PARA has an advantage in capturing antibody sequence information. Discussion: The antibody latent representation provided by PARA can substantially facilitate studies in relevant areas. We believe that PARA's superior performance in capturing antibody sequence information offers significant potential for both antibody property prediction and the development of novel therapeutics. PARA is available at https://github.com/xtalpi-xic.

High-content analysis (HCA) holds enormous potential for drug discovery and research, but widely used methods can be cumbersome and yield inaccurate results. Noise and high similarity in cell images impede the accuracy of deep learning-based image analysis. To address these issues, we introduce More Is Different (MID), a novel HCA method that combines cellular experiments, image processing, and deep learning modeling. MID effectively combines the convolutional neural network and Transformer to encode high-content images, effectively filtering out noisy signals and characterizing cell phenotypes with high precision. In comparative tests on drug-induced cardiotoxicity and mitochondrial toxicity classification, as well as compound classification, MID outperformed both DeepProfiler and CellProfiler, which are two highly recognized methods in HCA. We believe that our results demonstrate the utility and versatility of MID and anticipate its widespread adoption in HCA for advancing drug development and disease research.

ATP-sensitive potassium channels (KATP), composed of Kir6 and SUR subunits, convert the metabolic status of the cell into electrical signals. Pharmacological activation of SUR2-containing KATP channels by class of small molecule drugs known as KATP openers leads to hyperpolarization of excitable cells and to vasodilation. Thus, KATP openers could be used to treat cardiovascular diseases. However, where these vasodilators bind to KATP and how they activate the channel remains elusive. Here, we present cryo-EM structures of SUR2A and SUR2B subunits in complex with Mg-nucleotides and P1075 or levcromakalim, two chemically distinct KATP openers that are specific to SUR2. Both P1075 and levcromakalim bind to a common site in the transmembrane domain (TMD) of the SUR2 subunit, which is between TMD1 and TMD2 and is embraced by TM10, TM11, TM12, TM14, and TM17. These KATP openers synergize with Mg-nucleotides to stabilize SUR2 in the NBD-dimerized occluded state to activate the channel.

The RBD (receptor binding domain) of the SARS-CoV-2 virus S (spike) protein mediates viral cell attachment and serves as a promising target for therapeutics development. Mutations on the S-RBD may alter its affinity to the cell receptor and affect the potency of vaccines and antibodies. Here we used an in silico approach to predict how mutations on RBD affect its binding affinity to hACE2 (human angiotensin-converting enzyme2). The effect of all single point mutations on the interface was predicted. SPR assay results show that 6 out of 9 selected mutations can strengthen binding affinity. Our prediction has reasonable agreement with the previous deep mutational scan results and recently reported mutants. Our work demonstrated the in silico method as a powerful tool to forecast more powerful virus mutants, which will significantly benefit the development of broadly neutralizing vaccine and antibody.

ATP-sensitive potassium channels (KATP) are metabolic sensors that convert the intracellular ATP/ADP ratio to the excitability of cells. They are involved in many physiological processes and implicated in several human diseases. Here we present the cryo-EM structures of the pancreatic KATP channel in both the closed state and the pre-open state, resolved in the same sample. The nucleotides bind at the inhibitory sites of the Kir6.2 channel in the closed state but not in the pre-open state. Structural comparisons reveal the mechanism for ATP inhibition and Mg-ADP activation, two fundamental properties of KATP channels. Moreover, the structure also uncovers the activation mechanism of diazoxide-type KATP openers.

Background: HER2 transmembrane domain (TMD) mutation has been reported as a rare driver mutation associated with advanced stage disease and a poor prognosis in patients with lung adenocarcinoma (LUAD). We aimed to comprehensively profile the genetic landscape and treatment response information of HER2 TMD-mutant LUAD. Methods: An in-house database of 7,812 LUAD patients was screened for mutation prevalence. A multi-center cohort of 16 HER2 V659E-mutant patients and an external cohort of 38 HER2-mutant patients from cBioPortal with overall survival (OS) data were analyzed. Eight patients from the in-house cohort were included in the real-world study of treatment response. Molecular docking simulation and binding affinity prediction were performed. Results: In Chinese LUAD, the prevalence of HER2 TMD mutation was 0.18% (14/7,812), and 0.14% (11/7,812) for the HER2 V659E mutation. The most recurrent co-alteration was TP53 mutation (n=4, 25%) and HER2 amplification (n=2, 12.5%). TMD-mutant patients were diagnosed at more advance stages (P<0.001) and had poorer OS (median OS 10.0 vs. 61.6 months, HR =7.9, 95% CI: 1.0-61.0, P<0.001) than non-TMD mutations. The overall response rate of targeted therapy, chemo-based therapy, and immunotherapy was 57.1%, 22.2%, and 0%, respectively. We postulated to challenge the resistance of tyrosine kinase inhibitor (TKI) with another with stronger binding energy to HER2 and supported the conclusion with a successful case. Additionally, we demonstrated a three-month response to the off-label use of pyrotinib in fifth-line therapy. Conclusions: Comapred with non-TMD mtuations, HER2 TMD mutation is a rare driver mutation with poorer prognosis in LUAD. Targeted therapy is the dominant choice for patients harboring this targetable mutation and longer OS could possibly be achieved through rechallenge with TKI of stronger binding affinity. Response to fifth-line pyrotinib was observed.

While Bayesian Optimization (BO) has emerged as sample-efficient optimization method for accelerating drug discovery, it has rarely been applied to the process optimization of pharmaceutical manufacturing, which traditionally has relied on human-intuition, along with trial-and-error and slow cycles of learning. The combinatorial and hierarchical complexity of such process control also introduce challenges related to high-dimensional design spaces and requirements of larger scale observations, in which BO has typically scaled poorly. In this paper, we use penicillin production as a case study to demonstrate the efficacy of BO in accelerating the optimization of typical pharmaceutical manufacturing processes. To overcome the challenges raised by high dimensionality, we apply a trust region BO approach (TuRBO) for global optimization of penicillin yield and empirically show that it outperforms other BO and random baselines. We also extend the study by leveraging BO in the context of multi-objective optimization, allowing us to further evaluate the trade-offs between penicillin yield, production time, and CO$_2$ emission as by-product. Through quantifying the performance of BO across high-dimensional and multi-objective drug production optimization processes, we hope to popularize application of BO in this field, and encourage closer collaboration between machine learning and broader scientific communities.

Active fluids such as swarming bacteria and motile colloids exhibit exotic properties different from conventional equilibrium materials. As a peculiar example, a spherical tracer immersed inside active fluids shows an enhanced translational diffusion, orders of magnitude stronger than its intrinsic Brownian motion. Here, rather than spherical tracers, we investigate the diffusion of isolated ellipsoids in a quasi-two-dimensional bacterial bath. Our study shows a nonlinear enhancement of both translational and rotational diffusions of ellipsoids. More importantly, we uncover an anomalous coupling between particles' translation and rotation that is strictly prohibited in Brownian diffusion. The coupling reveals a counterintuitive anisotropic particle diffusion, where an ellipsoid diffuses fastest along its minor axis in its body frame. Combining experiments with theoretical modeling, we show that such an anomalous diffusive behavior arises from the generic straining flow of swimming bacteria. Our work illustrates an unexpected feature of active fluids and deepens our understanding of transport processes in microbiological systems.

The functions and elasticities of the cell are largely related to the structures of the cytoskeletons underlying the lipid bilayer. Among various cell types, the red blood cell (RBC) possesses a relatively simple cytoskeletal structure. Underneath the membrane, the RBC cytoskeleton takes the form of a two-dimensional triangular network, consisting of nodes of actins (and other proteins) and edges of spectrins. Recent experiments focusing on the malaria-infected RBCs (iRBCs) show that there is a correlation between the elongation of spectrins in the cytoskeletal network and the stiffening of the iRBCs. Here we rationalize the correlation between these two observations by combining the wormlike chain model for single spectrins and the effective medium theory for the network elasticity. We specifically focus on how the disorders in the cytoskeletal network affect its macroscopic elasticity. Analytical and numerical solutions from our model reveal that the stiffness of the membrane increases with increasing end-to-end distances of spectrins, but has a nonmonotonic dependence on the variance of the end-to-end distance distributions. These predictions are verified quantitatively by our atomic force microscopy and micropipette aspiration measurements of iRBCs. The model may, from a molecular level, provide guidelines for future identification of new treatment methods for RBC-related diseases, such as malaria infection.

The macroscopic properties, the properties of individual components, and how those components interact with each other are three important aspects of a composited structure. An understanding of the interplay between them is essential in the study of complex systems. Using axonal cytoskeleton as an example system, here we perform a theoretical study of slender structures that can be coarse-grained as a simple smooth three-dimensional curve. We first present a generic model for such systems based on the fundamental theorem of curves. We use this generic model to demonstrate the applicability of the well-known worm-like chain (WLC) model to the network level and investigate the situation when the system is stretched by strong forces (weakly bending limit). We specifically studied recent experimental observations that revealed the hitherto unknown periodic cytoskeleton structure of axons and measured the longitudinal fluctuations. Instead of focusing on single molecules, we apply analytical results from the WLC model to both single molecule and network levels and focus on the relations between extensions and fluctuations. We show how this approach introduces constraints to possible local dynamics of the spectrin tetramers in the axonal cytoskeleton and finally suggests simple but self-consistent dynamics of spectrins in which the spectrins in one spatial period of axons fluctuate in-sync.

Malaria is one of the most widespread and deadly human parasitic diseases caused by the Plasmodium (P.) species with the P. falciparum being the most deadly. The parasites are capable of invading red blood cells (RBCs) during infection. At the late stage of parasites' development, the parasites export proteins to the infected RBCs (iRBC) membrane and bind to receptors of surface proteins on the endothelial cells that line microvasculature walls. Resulting adhesion of iRBCs to microvasculature is one of the main sources of most complications during malaria infection. Therefore, it is important to develop a versatile and simple experimental method to quantitatively investigate iRBCs cytoadhesion and binding kinetics. Here, we developed an advanced flow based adhesion assay to demonstrate that iRBC's adhesion to endothelial CD36 receptor protein coated channels is a bistable process possessing a hysteresis loop. This finding confirms a recently developed model of cell adhesion which we used to fit our experimental data. We measured the contact area of iRBC under shear flow at different stages of infection using Total Internal Reflection Fluorescence (TIRF), and also adhesion receptor and ligand binding kinetics using Atomic Force Microscopy (AFM). With these parameters, we reproduced in our model the experimentally observed changes in adhesion properties of iRBCs accompanying parasite maturation and investigated the main mechanisms responsible for these changes, which are the contact area during the shear flow as well as the rupture area size.

We use high-speed X-ray phase-contrast imaging, weakly nonlinear analysis and boundary integral simulations to characterize the final stage of underwater bubble break-up. The X-ray imaging study shows that an initial azimuthal perturbation to the shape of the bubble neck gives rise to oscillations that increasingly distort the cross-section shape. These oscillations terminate in a pinch-off where the bubble surface develops concave regions that contact similar to what occurs when two liquid drops coalesce. We also present a weakly nonlinear analysis that shows that this coalescence-like mode of pinch-off occurs when the initial shape oscillation interferes constructively with the higher harmonics it generates and thus reinforce each other's effects in bringing about bubble break-up. Finally we present numerical results that confirm the weakly nonlinear analysis scenario as well as provide insight into observed shape reversals. They demonstrate that when the oscillations interfere destructively, a qualitatively different mode of pinch-off results where the cross-section profile of the bubble neck develops sharply-curved regions.

The presence of slight azimuthal asymmetry in the initial shape of an underwater bubble entirely alters the final breakup dynamics. Here, I examine the influence of initial asymmetry on the final breakup by simulating the bubble surface evolution as a Hamiltonian evolution corresponding to an inviscid, two-dimensional, planar implosion. I find two types of breakups: a previously reported coalescence mode in which distant regions along the air-water surface curve inwards and eventually collide with finite speed, and a hitherto uncharacterized cusp-like mode where the surface develops sharp tips whose radii of curvature are much smaller than the average neck radius. I present three sets of results that characterize the nature of this cusp mode. First, I show that the cusp-like mode corresponds to a phase space trajectory that passes close to a saddle-node structure. In other words, an evolution towards a cross-section shape with sharp tips invariably later evolves away from it. In phase space, this saddle-node separates coalescence modes whose coalescence planes lie along different spatial orientations. Second, I show that the formation of the sharp tips can be interpreted as a weakly first-order transition which becomes second-order, corresponding to the formation of a finite-time curvature singularity, in the limit that the initial perturbation amplitude approaches zero. Third, I show that, as the curvature singularity is approached, the maximum surface curvature diverges approximately as (tc − t)−0.8, where tc is the onset time of the singularity and the maximum velocity diverges approximately as (tc − t)−0.4. In practice, these divergences imply that viscous drag and compressibility of the gas flow, two effects not included in my analysis, become significant as the interface evolves towards the curvature singularity.

The disconnection of an underwater bubble illustrates how slight initial asymmetries can prevent the formation of a finite-time singularity. Creating a singularity by focusing a finite amount of energy dynamically into a vanishingly small amount of material requires that the initial condition be perfectly symmetric. In reality, imperfections are always present. We show a slight azimuthal asymmetry in the initial shape of the bubble neck excites vibrations that persist over time. As a result, the focusing singularity is generically preempted by a smooth contact.