324139 research outputs found
Sort by
Planting and MCMC sampling from the Potts Model
Get PDFWe consider the problem of sampling from the ferromagnetic q-state Potts model on the random d-regular graph with parameter β > 0. A key difficulty that arises in sampling from the model is the existence of a “metastability” window β ∈ (βu, β′ u), where roughly the distribution has two competing modes, the so-called disordered and ordered phases. This causes classical Markov-chain algorithms to be slow mixing from worst-case initialisations. Nevertheless, Helmuth, Jenssen and Perkins (SODA ’19) designed a sampling algorithm that works for all β, when d ≥ 5 and q = d Ω(d), using polymers and cluster expansion methods; more recently, their analysis technique has been adapted to show that a Markov chain (random-cluster dynamics) mixes fast when initialised appropriately, in the same regime of q, d, β.Despite these positive algorithmic results, a well-known bottleneck behind cluster-expansion arguments is that they inherently only work for large q, whereas it is widely conjectured that sampling on the random d-regular graph is possible for all q, d ≥ 3. The only result so far that applies to general q, d ≥ 3 is by Blanca and Gheissari who showed that the random-cluster dynamics mixes fast in the “uniqueness” regime β βc where we obtain an algorithm for all d ≥ 3 and q ≥ (5d) 5 , improving significantly upon the previous range of q, d by Carlson, Davies, Fraiman, Kolla, Potukuchi, and Yap (FOCS’22)
Activating molecules and complexes for diagnosis and therapy
Get PDFDevelopment of new molecules for cancer imaging, for diagnosis and prognosis evaluation, is crucial to expand the information available to clinicians. The formation of activatable profluorophores is desirable to enable imaging of disease and upregulation of enzymes or biological molecules in certain regions, such as in the presence of hypoxia (low oxygen concentration) in cancerous tissues.In hypoxic tissue, found in solid tumours, the effects of both chemotherapy and radiotherapy are reduced, therefore selective delivery of a compound which is activated in the presence of both hypoxia and X-ray irradiation is a route to overcome this. Chapter II highlights the design of a series of profluorophores, based on known fluorophores and the tetrafluoroaryl azide group, as proof-of-concept work to determine whether activation of a profluorophore under hypoxia and X-ray irradiation is orthogonal or can be used advantageously in combination. The chapter explores profluorophore design and the poor activation of the novel profluorophores under reductive chemical and enzymatic conditions and X-ray irradiation, highlighting the need for development of new moieties which can be activated at lower, clinically relevant doses of X-ray irradiation to further investigate the orthogonality of activation.Investigation of lanthanide complexes as imaging agents, due to their advantages over organic fluorophores, is an expanding field and these complexes have utility for both optical imaging and MRI, based on the lanthanide metal employed (Eu, Tb or Gd). Chapter III discusses the synthesis, spectroscopy and activation of complexes which contain ester linkages to common hypoxia-activated groups and highlights the change in luminescence and relaxivity between the profluorophores and known LnDOTA complexes. It is proposed that these complexes release LnDOTA in the presence of chemical and enzymatic reductants via a hydrolysis mechanism, therefore they do not exhibit the desired selectivity profile for biological applications and emphasise the need to balance activity and selectivity. However, these novel probes act as a proof-of-concept for activatable optical imaging and MRI complexes and give insight into the design considerations needed for their future development.Chapter IV builds on the drawbacks from chapter III, involving the design of leucine and alkyl appended azobenzene bimetallic lanthanide complexes, aiming for improved cell permeability and emission intensity for biological application of reductively activated lanthanide complexes in vivo. However, derivatised azobenzene bimetallic complexes were synthetically challenging to make and further work is needed to expand on approaches to overcome the problem of cell permeability.Chapter V summarises the work carried out in this thesis and chapter VI describes the experimental synthetic and assay procedures utilised to obtain the results in chapters II-IV
Optimal inter-electrode distances for maximizing single unit yield per electrode in neural recordings
Get PDFState-of-the-art high-density multielectrode arrays enable the recording of simultaneous spiking activity from hundreds of neurons. Although significant efforts have been dedicated to enhancing neural recording devices and developing more efficient sorting algorithms, there has been relatively less focus on the allocation of microelectrodes–a factor that undeniably affects spike sorting effectiveness and ultimately the total number of detected neurons. Here, we systematically examined the relationship between optimal electrode spacing and spike sorting efficiency by creating virtual sparser layouts from high-density recordings through spatial downsampling. We assessed spike sorting performance by comparing the quantity of well-isolated single units per electrode in sparse configurations across various brain regions (neocortex and thalamus), species (rat, mouse, and human) and various spike-sorting algorithms. Enabling the theoretical estimation of optimal electrode arrangements, we complement experimental results with a geometrical modeling framework. Contrary to the general assumption that higher electrode density inherently leads to more efficient sorting, both our theoretical and experimental results reveal a clear optimum for electrode spacing specific to species and regions. We demonstrate that carefully choosing optimal electrode distances could yield a total of 1.7–3.75 times increase in spike sorting efficiency. These findings emphasize the necessity of species- and region-specific microelectrode design optimization
Light-induced analgesia provides a drug-free optical method for pain relief via activation of TRAAK k + channels
Get PDFPain management in animal experimentation is crucial for both ethical and scientific reasons, as unmanaged pain can distort physiological responses compromising data reliability. Current strategies are often invasive and pharmacology-based, introducing variability and confounding effects. Here, we present Light-Induced Analgesia, a drug-free, non-invasive method for pain relief in animals. We show that 365 nm illumination activates the pain-inhibitory TRAAK two-pore domain potassium (K2P) channel. This activation is driven by the oxidation of a native methionine at TRAAK’s regulatory fenestration site, triggering a conformational switch from its inactive (down) to active (up) state. We further demonstrate that this mechanism can be transferred to other related K2Ps via a single-point mutation, rendering them light-sensitive. In rodents, gentle skin exposure to 365 nm is sufficient to activate endogenous TRAAK, silence nociceptors, and produce potent, long-lasting analgesia that outperforms standard treatments. Light-Induced Analgesia thus offers an effective, drug-free alternative that can enhance animal welfare and experimental reliability in preclinical research
Samanta Schweblin. El buen mal. Buenos Aires: Penguin Random House, March 2025. ISBN 978-987-769-389-8
No full textPre-COVID-19 ex vivo cross-reactive IFN-γ cellular response to SARS-CoV-2 spike overlapping peptides is more prevalent among Kenyan compared to Swedish adults
Get PDFBackground: Global WHO data indicate that Sub-Saharan African (SSA) countries, such as Kenya, experienced reduced coronavirus disease 2019 (COVID-19) severe-morbidity and mortality burdens relative to their more affluent counterparts in Europe, Asia, and North America. Methods: We analysed peripheral blood mononuclear cells (PBMC) samples collected from Kenya and Sweden before and during COVID-19. Pre-COVID-19 samples were available for 80 adults and 10 infants from Kenya, and 20 adults from Sweden. COVID-19 samples were available for 39 Kenyan adults. The samples were analysed for ex vivo IFN-γ secretion using an Enzyme-Linked Immunosorbent (ELISpot) assay following in vitro stimulations with overlapping SARS-CoV-2 spike-protein peptides. T-cells expressing IFN-γ, IL-2, TNF-α, CD154, and CD107a were assessed following similar stimulations, using intracellular cytokine staining (ICS) and multiparameter flow cytometry. Results: 55.7% of the Kenyan pre-COVID-19 adult samples were classified as responders by ELISPOT responses to spike-protein peptides, compared with 28% of Swedish pre-COVID-19 adult sample (p = 0.04). The frequencies for SARS-CoV-2 spike-specific TNF-α CD4+, TNF-α CD8 + and IFN-γ CD8 + T-cell responses, tended to be higher in the Kenyan adults although these differences did not reach statistical significance. Conclusion: Pre-COVID-19 T-cell responses could contribute to lower morbidity and mortality associated with SARS-CoV-2 infections in SSA relative to Europe, Asia, and North America
Tumours in the newborn - a systematic review and review of grey literature exploring differences in the types, treatments, and outcomes of neonatal tumours in high income versus low-middle income countries
Get PDFNeonatal Tumours are one of the least researched conditions in paediatrics and pose a significant surgical challenge due to the lack of accessibility to and standardization of treatment, especially in lower income regions of the world. Combining a systematic review and a review of grey literature, this research project explores how the types, treatments, outcomes, and parental perspectives of Neonatal Tumours differ across income levels, and it highlights a lack of literature in Low-Middle Income Countries. The systematic review aims to examine existing literature regarding Neonatal Tumours. Thirty- four reviews were deemed eligible. These studies collectively reported on 2023 neonatal patients, with 1759 of these patients from High Income Level Countries, 178 of these patients from Upper Middle Income Level Countries, 86 of these patients from Lower Middle Income Level Countries and 0 from Low Income Level Countries. Overall Survival was 79.3%: 79.1% in High Income Level Countries, 71.9% in Upper Middle Income Level Countries, and 98.8% in Lower Middle Income Countries. These differences were found to be statistically significant. However, it is important to note the disparities in available patient data between country income levels. The lack of data in Low-Middle Income Level countries emphasizes a need for more research in these areas of the world. The review of Case Reports dives into the grey literature on this subject, providing observations on types of tumours, demographics, and outcomes by country income level and discussing the importance of Case Reports in research. Using a novel system of categorization, case reports were sorted into five separate categories based on what each intends to present: Management suggestions/dilemmas (n=31), Diagnostic suggestions/dilemmas (n=16), report of a Unique Presentation of a case (n=12), report of a Rare Case (n=8), and Etiology / Pathogenesis of the disease (n=4)
Can editing LLMs inject harm?
No full textLarge Language Models (LLMs) have emerged as a new information channel. Meanwhile, one critical but underexplored question is: Is it possible to bypass the safety alignment and inject harmful information into LLMs stealthily? In this paper, we propose to reformulate knowledge editing as a new type of safety threat for LLMs, namely Editing Attack, and conduct a systematic investigation with a newly constructed dataset EditAttack. Specifically, we focus on two typical safety risks of Editing Attack including Misinformation Injection and Bias Injection. For the first risk, we find that editing attacks can inject both commonsense and long-tail misinformation into LLMs, and the effectiveness for the former one is particularly high. For the second risk, we discover that not only can biased sentences be injected into LLMs with high effectiveness, but also one single biased sentence injection can degrade the overall fairness. Then, we further illustrate the high stealthiness of editing attacks. Our discoveries demonstrate the emerging misuse risks of knowledge editing techniques on compromising the safety alignment of LLMs and the feasibility of disseminating misinformation or bias with LLMs as new channels
Exclusionary government rhetoric and migration intentions
Get PDFWe investigate whether exclusionary government rhetoric targeting a minority group affects residents’ migration decisions. In 2019, almost 100 local governments in Poland voted to declare their localities "free from LGBTQ ideology," providing a unique setting in which government narratives suddenly changed, but the legal situation of targeted minorities remained the same. We study the impact of these declarations on migration intentions using novel data on domestic and international job search from a large global job site. Comparing counties with anti-LGBTQ resolutions to neighboring control counties in a difference-in-differences design, we find that the resolutions increased domestic out-of-county job search by 12 percent and international job search by 15 percent. Our results are likely driven by the shock to beliefs about local social norms, as we find the largest effects in counties with relatively low prior support for far-right parties. We also present suggestive evidence that the rise in job search translated into actual migration, with the treated counties losing nearly 1 percent of their young adult population