21793 research outputs found
Sort by
Molecular Phylogenetic Analysis of Woody Plants in India and Sri Lanka
No full textThe Indian subcontinent, encompassing India and Sri Lanka, is a region of exceptional plant diversity shaped by deep geological and evolutionary history. Its forest communities, dominated by woody plants, harbour a large proportion of this biodiversity. Understanding the evolutionary relationships among these taxa is crucial for exploring the processes of speciation, extinction, dispersal, and assembly of ecological communities.
This study reconstructs a phylogenetic framework for 2,208 woody plant species spanning 47 orders and 143 families using five commonly used genetic markers: chloroplast genes or genomic regions (rbcL, matK, psbA-trnH) and nuclear ribosomal genomic regions (ITS1, ITS2). A supermatrix alignment of nucleotide sequences from 1,356 taxa was generated and used for multilocus-based reconstruction of the phylogeny. The inclusion of nuclear markers notably enhanced tree resolution, reducing overall polytomies from 13.48% to 6.63%. For example, polytomies in Ficus decreased from 67.57% to 20%, and Cinnamomum nodes were fully resolved where ITS data were present.
Comparisons between chloroplast- and nuclear DNA sequence data-based trees revealed lineage-specific variation in phylogenetic informativeness. Nuclear ITS data improved resolution in Laurales, Malpighiales, Rosales, and Myrtales, while chloroplast markers performed better in Lamiales, Ericales, and Brassicales.
These results provide a robust phylogenetic framework for understanding community assembly and diversification processes in the forests of India and Sri Lanka, offering valuable insights for ecological research and biodiversity conservation
Towards High-Quality Clustering and Analysis Algorithms for Categor- ical Data
No full textCategorical data are prevalent across various disciplines, making clustering a valuable tool for analysis. However, clustering categorical data is particularly challenging due to its non-numerical nature and multi-modal distributions. While numerous techniques have been developed to address these challenges, several issues persist. Since most of the proposed techniques adapt algorithms like k-modes, which were originally designed for categorical data, they often fail to fully exploit and capture unique characteristics of categorical datasets. A main problem with those techniques require setting the number of clusters, which restricts their application and may result in bias when no expert knowledge is available. Additionally, since they tend to choose cluster seeds randomly, they perform well on binary-class data but struggle when applied to multi-class or imbalanced binary datasets. In this thesis, we propose three techniques specifically designed to address these challenges.
First, we introduce a cohesion-based clustering process that determines the potential number of clusters dynamically and that also allows detection of small clusters without relying on k-means or k-modes-like methods. Unlike conventional clustering algorithms that assign weights to all the attributes, we adopt a mechanism that assigns weights to clusters at the attribute-value level, improving cluster cohesion and interpretability. Second, we develop multi-criteria subspace-based clustering techniques that unlike the traditional subspace solution, search the entire space. Our techniques leverage two existing clustering strategies, namely density-based and theoretical clustering, to identify small, non-redundant clusters. To ensure effective merging, we extend the hierarchical clustering technique that allows discovered clusters to be combined while preventing small clusters from being absorbed into larger ones. Third, we study measuring quality and diversity methods in ensemble clustering selection. Existing methods often rely on external validation metrics, which are not well-suited to the characteristics of categorical data. Those metrics tend to introduce redun- dancy and favour large clusters, potentially overlooking smaller yet meaningful ones. Furthermore, most current approaches assess quality and diversity only at the cluster level, neglecting more gran- ular and informative perspectives. To overcome these limitations, we adopt and extend a measure based on granular computing, which aligns more naturally with the discrete and multi-level struc- ture of categorical data. The result helps evaluate quality and diversity at the class level as well as the object level, enabling a more nuanced and effective ensemble selection process. To evaluate its performance, we conduct extensive experiments using real-world and synthetic benchmark datasets. The experiment results and their analyses demonstrate overall enhanced clustering performance and helped identify small clusters in certain datasets
Follower Count or Expertise? Cracking the Influencer Code for Startups
No full textThis study investigates the impact of influencer type in digital marketing, specifically examining how follower size and perceived expertise affect consumer outcomes, such as attitude, engagement, and purchase intention, in the context of a utilitarian product. It also explores how brand type (established vs. start-up) moderates these relationships and whether perceived trust in the influencer mediates them. While influencer marketing is widely used, most existing research focuses on hedonic products. Little is known about how influencer type interacts with brand type to shape consumer attitudes and behaviors toward utilitarian products. To address this gap, two experimental studies were conducted. Study 1 examined the interaction between influencer follower size (mega vs. micro) and brand type (established vs. start-up) on consumer responses. Study 2 explored the interaction between influencer expertise (expert vs. lifestyle) and brand type, while also testing the mediating role of perceived trust in the influencer. The findings reveal that influencer effectiveness varies depending on follower size, expertise, and brand type. Trust in the influencer significantly mediates the effects on consumer attitudes and intentions. Theoretically, this study extends influencer marketing research into utilitarian contexts. Practically, it provides guidance for marketers, particularly those in start-ups, on selecting appropriate influencer types based on brand and product characteristics
Exploring Electrophysiological Changes in Adolescents With and Without Concussion Across the First Month After Injury
No full textConcussion recovery is primarily based on the full resolution of symptoms, but neurophysiological changes are present and may persist beyond clinical recovery. Our objectives were to 1) compare the evolution of electrophysiological and clinical outcomes between adolescents with and without concussion over the first month after injury and 2) determine the relationship between electrophysiological and clinical outcomes. Adolescents with concussion (n=53) were assessed within 10 days of and at 30-days post-injury, while adolescents without concussion (n=15) were tested at arbitrary dates but with a similar interval between sessions. Resting-state electroencephalography (EEG), Post-Concussion Symptom Inventory, Pediatric Quality of Life Inventory, and Revised Children’s Depression and Anxiety Scale were completed at all study visits. Separate mixed-effects models evaluated the effect group, time, and their interaction on clinical and EEG (F3 delta power, F8 delta power, P4 delta power, T6 beta power, and O1 beta power) variables, with all models controlling for age and sex. Pearson’s correlations examined the associations between EEG power and clinical symptoms, separately for each group and study visit. Results revealed that clinical measures improved over time (primarily driven by changes in the concussion group); however, no significant effects of group, time, or their interaction were observed for any EEG variable. Few significant correlations between EEG and clinical outcomes were observed. These results suggest that EEG outcomes do not change over the first month following concussion. Future studies should include larger sample sizes and different EEG features, which may be more sensitive to changes over time after concussion
Evolutionary Strategies Against Antimicrobial Resistance
No full textAntibiotic resistance threatens to undo many of the advancements of modern medicine. A slow antibiotic development pipeline makes it impossible to outpace bacterial evolution, making alternative strategies essential to combat resistance. In this study, I use large scale experimental evolutions powered by the soft agar gradient evolution (SAGE) platform to investigate the evolutionary trade-offs associated with antibiotic resistance, and how they can be leveraged to combat the emergence of resistance.
The study begins with the finding that a chloramphenicol (CHL) resistant Escherichia coli (E. coli) mutant exhibits a markedly reduced rate of resistance evolution against other antibiotics. I show that this slow adaptation is linked to the fitness costs associated with resistance, which bacteria often readily overcome through compensatory evolution. Further screening identifies fitness costs which cannot be easily compensated for, highlighting an opportunity to exploit these trade-offs to slow down the emergence of resistance. However, the translatory potential of the findings from SAGE to the clinic remained unclear.
To test the utility and clinical relevance of SAGE, I first expand its applicability to a broader range of antibiotics by supplementing the evolution medium with xanthan gum. Xanthan gum is a water-binding polysaccharide that significantly reduces synaeresis of the agar-based medium, enhancing evolution in SAGE. To demonstrate its capacity to uncover resistance mechanisms I use this modified platform to characterize the evolution of resistance to the lipopeptide tridecaptin A1—an antibiotic previously thought to be impervious to resistance. I then assess the clinical applicability of the evolutionary trade-offs observed in SAGE-derived mutants by comparing outcomes from SAGE to those obtained using other widely used laboratory evolution platforms, as well as clinical bacterial datasets. These analyses reveal that SAGE more accurately reproduces clinically relevant patterns of fitness trade-offs than the alternative platforms tested. One such trade-off, collateral sensitivity (CS), has recently been proposed to be useful in mitigating resistance in sequential antibiotic therapies, where antibiotics are applied one after the other. But large-scale evolutionary studies to determine its role and effectiveness in sequential regimens were missing.
I use over 450 evolution experiments to test the role of CS in resistance mitigation in four proposed drug pairs. I find that resistance to both drugs evolves readily, and that collateral sensitivity does not hinder the evolution of multidrug resistance or promote resensitization. However, if resistance to drug B reduces susceptibility to A in an A-B drug sequence, a phenomenon I term backward CS, resistance to A can be reduced. As an example, I demonstrate that β-lactam resistant E. coli cells frequently lower their resistance to β-lactams upon aminoglycoside resistance acquisition due to conflicting modifications to the proton motive force and efflux pumps. This suggests that the levels of resistance evolved can be kept in check by leveraging backwards CS to resensitize cells as antibiotic resistance evolves. However, the levels of resensitization achieved were two-fold on average, often not sufficient to reduce resistance below clinical breakpoints.
Finally, I introduce sequential antibiotic regimens composed of three drugs or “tripartite loops” to contain resistance within a closed drug cycle. Through 424 discrete adaptive laboratory evolution experiments I show that as bacteria sequentially evolve resistance to the drugs in a loop, they continually trade their past resistance for fitness gains, reverting back to sensitivity via four-to-eight-fold reductions in resistance on average. Through fitness and genomic analyses, I find that tripartite loops guide bacterial strains towards evolutionary paths that mitigate fitness costs and reverse resistance to component drugs in the loops, driving levels of resensitization not achievable through previously suggested pairwise regimens. I then apply this strategy to reproducibly resensitize or eradicate four multidrug-resistant clinical isolates over the course of 216 evolutionary experiments. Resensitization occurred even when bacteria adapted through plasmid-bound mutations instead of chromosomal changes, showing the robustness of this strategy.
In conclusion, this work demonstrates that the evolutionary trade-offs accompanying antibiotic resistance can be strategically exploited to limit or reverse resistance evolution. I highlight the importance of studying the evolutionary aspect of antibiotic resistance to inform rational treatment strategies and restore efficacy of existing antibiotics. As the pace of novel antibiotic discovery continues to lag behind resistance evolution, such evolution-based approaches may be essential for extending the lifespan of our current antimicrobial arsenal
Hardening Fifth-Generation (5G) Network Slicing: Protecting User Data and Detecting Quality-of-Service (QoS)-Based Attacks
No full textFifth-generation (5G) and beyond networks are driven by the growing number of users and sophisticated services, as industries move toward greater automation. To ensure adequate user experience, specific Quality of Service (QoS) requirements, such as low latency and high reliability, must be defined. Consequently, network slicing has emerged as a key feature of 5G, allowing operators to provide customized logical networks, known as Network Slices (NSs), each tailored to its envisioned services. NSs adopt 5G’s cloud-native Service-Based Architecture (SBA), which relies on common virtualization technologies. Additionally, to maintain QoS guarantees in a Network Slice (NS), the Third Generation Partnership Project has described a QoS model with QoS monitoring and reporting procedures. Despite their advantages, both SBA and QoS monitoring and reporting extend 5G’s attack surface. The former permits the leakage of sensitive User Equipment (UE) data (e.g., location), and the latter can be abused by a UE falsifying trusted measurements over the user plane to generate excessive reporting on the control plane, staging a novel QoS-based Attack (QoSA). In this thesis, we focus on securing 5G network slicing against inter-slice attacks that violate the Service Level Agreements for legitimate UEs served by NSs sharing the affected resources. Thus, we first disclose a prevention technique for protecting 5G against data theft attacks. Then, we introduce our novel attack, QoSA, and evaluate its impact using our testbed. This testbed was built by extending the open-source free5GC code to support standardized 5G QoS monitoring and reporting. Our experiments demonstrated that QoSA can lead to disruptions that affect service delivery. Accordingly, we present QoSA-Officer, an Attention-based Long-Short-Term Memory Autoencoder (Attention LSTM-AE) for detection, trained on QoS-relevant features from our emulations, enabling the distinction between QoSA and native QoS interruptions
Daily Proof of Liabilities
No full textA proof of solvency’s goal is to demonstrate that a cryptocurrency exchange possesses sufficient funds to satisfy client withdrawals. In this thesis, we introduce an improvement to the
prevailing way of building a proof of liabilities. We use the Nova novel way of proving that a
balance is included in the proof of liabilities (i.e. proof of inclusion), and apply it to the proof of
liabilities itself. We use the circuit designed to show the proof of inclusion of a Merkle tree, and
modify it to prove a list of balance changes in the Merkle tree. While this is slower than producing
the whole Merkle tree when you have many changes, this new circuit design enables to separate
the proof into multiple smaller proofs, enabling the use of the Nova folding scheme. The folding
of arithmetics circuits reduces the computation needed for a daily proof of liabilities, enabling the
possibility of obtaining this proof at a higher frequency, potentially as frequently as every block
Modification affordances: user-modified objects as inspiration for evolving industrial design practices
No full textCollapsed into every mundane, everyday object, that surrounds us are complex, mysterious, and interscalar worlds. As we begin to unpack those worlds we develop the understanding that our everyday objects are anything but mundane; rather they carry within them design history, cultural context, the marks of our systems of production, social biases, and possibilities. Using scalar thinking we can expand our perceptions of these everyday objects and open a discussion about who is most often failed by the shifts in scale required for mass produced objects to be designed, manufactured, and used. I propose that by looking at examples of where design has failed, we can learn new ways of designing to minimize exclusionary features.
Examples of user-modified objects were collected from study participants, and interviews were conducted about their relationships to modified objects, and the motivations behind their modifications. The qualitative coding of these interviews suggests reasons why people typically modify objects, considerations for modifications, and insight into the user-modification process. A research-creation process follows, focused on one of the participants' needs for a better grip for the Nintendo Switch. This custom modification process highlights the presence of design elements that allow for adaptations to take place. I propose that these design elements be considered “modification affordances”. Modification affordances can be seen as a point-in-time solution to consider how we might shift production away from mass produced markets and towards individual and custom alternatives, which can serve to benefit both people and the planet
Assessing L2 Pronunciation with Automatic Speech Recognition Dictation Tools
No full textThe use of automatic speech recognition (ASR) to score pronunciation placement tests offers language institutions an efficient alternative to human raters, addressing common challenges such as rater reliability and high labour costs. However, the customizable ASR systems used to create scoring models are costly, making them unaffordable for most institutions. As an alternative, this dissertation explored the feasibility of using transcripts from Google Voice Typing (GVT), a free and readily available dictation-based software, to provide automated scores for pronunciation assessments. Via two empirical studies, it addressed the following overarching research question “What are the affordances offered by dictation ASR in an L2 pronunciation assessment context?”
In the first study (Manuscript A), human-rated and GVT-rated scores of 56 pronunciation placement tests were compared, showing strong correlations. However, when the samples were divided by proficiency levels, there were weak correlations for high-proficiency users, raising concerns about the reliability of the test scores. To explain this finding, it was hypothesized that some high-proficiency test takers received low GVT scores due to problematic linguistic elements in some test items (e.g., highly infrequent words, unusual collocations). Overall, this study showed that scoring pronunciation tests with GVT is feasible, with the caveat that reliability issues need to be explored to ensure test validity and reliability.
As a follow-up, the second study (Manuscript B) explored the effect of word frequency, unusual collocations, and phonologically ambiguous items on GVT transcription accuracy, with the aim of supporting the design of valid and reliable pronunciation tests. Four highly intelligible English speakers recorded 60 sentences targeting these three features. The recordings were transcribed by GVT and scored for accuracy, while eight human raters carried out an intelligibility transcription task which was also scored for accuracy. For GVT, the results suggest that lower-frequency vocabulary and phonologically ambiguous phrases were particularly challenging, while sentences containing names, proper nouns, or unusual collocations were almost always accurately transcribed. In contrast, transcriptions produced by human raters showed great variability and tended to be less accurate than those generated by GVT. These results indicate that certain features are difficult for both human raters and GVT to transcribe, even when produced by highly intelligible speakers. This highlights the importance of careful task design to avoid features that may compromise transcription accuracy. To ensure valid, reliable scores and fair decision-making, transcription accuracy must be verified through systematic test piloting.
The findings of this dissertation emphasize that GVT can be used to develop cost-effective scoring models for pronunciation placement tests, while also highlighting the need for careful task design to avoid the use of language that may compromise transcription accuracy and unfairly penalize test takers. These results also offer practical implications for classroom-based pronunciation assessments
Patient-Facing Application design: A User-Centred, FMEA & SIPOC Framework for Resilient Diagnostic Workflows
No full textThis dissertation reports the results of analyzing the possible failures in medical test ordering and delivery (MTOD) systems within McGill University Health Center (MUHC) in Quebec. In conjunction with an external consultant and the Fédération des médecins spécialistes du Québec, we have mapped the daily test ordering and delivery process of several medical departments within this hospital. All sessions were video‑recorded, fully transcribed, and iteratively coded. Each transcript was decomposed into discrete task
blocks from which individual SIPOC diagrams were prepared. Read side‑by‑side, the SIPOCs exposed hidden hand‑off gaps and bottlenecks. For every high‑risk step a Failure‑Mode‑and‑Effects‑Analysis (FMEA) table was created. The combined SIPOC & FMEA guided the sprint that produced a patient‑facing pilot application: a stand‑alone layer that sits beside OACIS, neutralizing the three most‑frequent failure modes, poor communication
of test results, missed bookings, and bookings that were never created. Preliminary analysis indicates that integrating SIPOC mapping and FMEA within a user‑centered design workflow streamlined decision‑making and enabled rapid prototyping of the pilot MTOD application. However, additional validation is needed before any claims about risk reduction or clinical impact can be made