324139 research outputs found
Sort by
Novel drinkable oxygen nanobubbles: a pilot randomised, double-blind, cross-over trial on pulmonary fibrosis patients
No full textThe impact of artificial intelligence on venture capital: a critical outlook
Get PDFThis paper examines how the adoption of artificial intelligence (AI) technologies is transforming the venture capital industry. Leveraging interviews with industry practitioners who are actively working on the adoption of AI tools, we develop a critical outlook on the likely changes awaiting the industry. We find that AI tools accelerate the sourcing and due diligence of venture deals. However, the final authority to make investment decisions remains with humans. We identify socially grounded conviction and gut feelings as two key human proficiencies that AI systems struggle to replicate. Relationships embedded in their professional networks also allow venture capitalists to provide value-added support to their founders in ways that cannot be easily replaced by AI systems. As AI systems are adopted industry-wide, they will broaden founders’ access to funding and shift market power away from investors. It is not the AI infrastructure itself, but human proficiencies, such as socially grounded conviction, gut feeling, and networks, that will allow venture capital firms to competitively differentiate themselves
Menin maintains enhancer-promoter interactions in a leukemia-specific manner
Get PDFInhibition of the protein-protein interaction between Mixed Lineage Leukemia (MLL/KMT2A) protein and the adapter protein Menin is a promising therapy for both high-risk MLL-rearranged and NPM1-mutant (NPM1c) acute leukemias. However, the exact function of Menin in transcription regulation remains unclear, limiting our understanding of the utility of Menin inhibitors. Here, we compared the transcriptional responses of MLL-AF4 and NPM1c leukemias to Menin inhibition. We observed broad, acute transcriptional dysregulation in MLL-AF4 cells, whereas NPM1c showed few targets directly regulated by Menin binding. Screening for Menin interacting proteins by co-immunoprecipitation followed by mass spectrometry revealed a greater enrichment for transcriptional regulators and elongation factors in MLL-AF4 cells compared to NPM1c cells, pointing to protein complex and cell context-specific activity. Finally, we used high-resolution Micro-Capture-C to demonstrate that Menin regulates intragenic enhancer activity and maintains enhancer-promoter contacts in MLL-AF4 cells but not in NPM1c cells. Together, our findings show that Menin exerts distinct transcriptional effects depending on the protein complex it associates with, which is key to understanding how patients respond transcriptionally to Menin inhibition in distinct disease contexts
Associations between demographic, clinical and dietary factors and flares in inflammatory bowel disease: the PRognostic effect of Environmental factors in Crohn’s and Colitis (PREdiCCt) prospective cohort study
Get PDFBackground: IBD is characterised by recurrent flares, but evidence on whether modifiable dietary factors influence flare risk is limited. Objective: The PREdiCCt study was designed to examine demographic, clinical and dietary factors associated with disease flare among patients with IBD in self-reported remission. Design: Multicentre, prospective cohort study conducted across 47 UK centres. Patients with Crohn’s disease (CD), ulcerative colitis (UC) or IBD unclassified (IBDU) in self-reported remission were prospectively followed up. The baseline diet was assessed using a validated food frequency questionnaire. The primary outcome was time to patient-reported flare (captured by monthly IBD-Control) and objective flare (clinical flare plus C-reactive protein >5 mg/L and/or faecal calprotectin (FC) >250 µg/g with treatment escalation). Associations were evaluated using Cox frailty models adjusted for demographic, clinical and biochemical variables, including baseline FC. Results: Between November 2016 and March 2020, 2629 participants (1370 CD; 1259 UC/IBDU) were enrolled and followed up for a median of 4.1 years (IQR 3.0–5.0). Baseline FC was strongly associated with patient-reported flares (FC ≥250 µg/g: adjusted HR (aHR) 2.22; FC 50–250 µg/g: aHR 1.52 (reference <50 µg/g)) and objective flares (FC ≥250 µg/g: aHR 3.25; FC 50–250 µg/g: aHR 1.98). In UC, higher total meat intake was associated with increased risk of objective flares (highest versus lowest quartile: aHR 1.95, 95% CI 1.07 to 3.56). No consistent associations were observed for ultraprocessed foods, fibre or polyunsaturated fatty acids and flare. Conclusion: Higher habitual meat intake was associated with increased risk of objective flare in UC, suggesting diet may contribute to flare susceptibility in specific patient groups. Trial registration number: NCT03282903
Prediction-powered machine learning for model selection and uncertainty
Get PDFThis thesis explores model selection and uncertainty through the lens of prediction. Building on recent developments in Bayesian predictive inference, we approach uncertainty as a missing data problem, extending the logic of the bootstrap by treating future observations as the basis for inference. This framework offers a complementary perspective to conventional frequentist and Bayesian methods, as it supports probabilistic uncertainty quantification without requiring the subjective specification of a prior distribution. We first apply this lens to model uncertainty and hypothesis testing, proposing a novel procedure where uncertainty is propagated via the recursive imputation of new data, using a one-step-ahead model selection criterion. We then broaden this view, arguing that a model’s sequential predictive behavior — specifically, its production of conditionally identically distributed updates — can be used to characterize its coherence, allowing for Bayesian-style uncertainty even in plug-in or frequentist settings. Finally, we apply predictive model ensembling to causal treatment effect estimation, introducing a random forest method that targets relative risk heterogeneity, and demonstrating its application to data from a major cardiovascular clinical trial. Taken together, these results argue that predictive resampling methods, grounded in bootstrap principles, can provide flexible and principled tools for model evaluation and validation
Loss Counterfactuals
Get PDFPrivate law uses counterfactual reasoning to determine if loss—for which compensatory damages may be awarded—has been suffered because of a wrong: is the claimant in a worse position than it would have been in without the wrong? Counterfactual reasoning is, therefore, an essential part of private law. Yet, it is far from clear what this counterfactual for determining loss is. Different counterfactuals seem to be used, depending on the claim and the claim’s circumstances. Counterfactuals remain, in general, an undertheorised concept. This article proposes a new, principled theory, and conceptualisation, of counterfactuals in private law. Its central claim is that, as a rule, the counterfactual for determining loss is, and should be, ‘what would have happened without the wrong’, but that there are, and should be, exceptions to this rule, if normatively justifiable
Geometry of smooth Gaussian fields
Get PDFGaussian fields are ubiquitous in probability as they are scaling limits of many natural objects, and in applied science as they are instrumental in modelling natural phenomena. In this thesis, we study geometric features of smooth Gaussian fields, like the measure of level sets and the structure of critical points of the field. Large-scale geometry, i.e. studying geometric observables in a domain where the domain size goes to infinity, is of particular interest.In the first chapter, we formally define smooth Gaussian fields and state their basic properties. Then we briefly explain connections to other topics in mathematics, including percolation theory, quantum chaos, and real algebraic geometry.Next, we study the measure of level sets of stationary Gaussian fields. Given two Gaussian fields that are close, we ask how close the measures of their level sets are in a given domain. Here we bring novel ideas (in this context) from geometric analysis to answer this question. We prove the convergence of the Hausdorff measures of level sets of smooth Gaussian fields when the levels converge. Given two stationary fields f_1,f_2 on a probability space , we estimate the difference of Hausdorff measures of level sets in expectation, in terms of C^2-fluctuations of the field F= f1−f2. We prove this using the mean curvature representation of the difference in measure of level sets. This approach is different from using a Kac-Rice type formula as the primary tool in the analysis. This chapter is based on joint work with Dmitry Belyaev [BH23]. We extend this result in several directions, thanks to a subsequent work of Peccati and Stecconi [PS25]. The extended result also applies to some non-stationary fields and improves the exponent of the bound to a (conjecturally) optimal one. Also, we require only C^0-fluctuations of the difference field F. The extended result is an ongoing work with Michele Stecconi and Francesca Pistolato.In the third chapter, we study the critical point structure of smooth Gaussian fields. We consider the point process (in R^d) of local maxima of Gaussian fields, with sufficient decay of correlation at infinity, above a level u. We show that this point process, rescaled appropriately, converges weakly to a Poisson point process in the limit u → ∞. In the literature, high excursion sets of many smooth and non-smooth Gaussian processes of decaying correlation are well studied [LLR83]. Also, for Gaussian processes with some Markov property, like Brownian motion or the Gaussian free field, high points (after suitable rescaling) have been shown to converge to a Poisson process. Our proof relies on the classical observation that simple point processes are characterised by avoidance probabilities (i.e. P(η(B) = 0) for Borel sets B). Then, we approximate avoidance probability with excursion probability, where the latter is well studied.Another significant result in the thesis is a quantified version of the Poisson convergence of high critical points of smooth Gaussian fields. We show that, for Bargmann-Fock field in two dimensions, the total variation distance between a Poisson random variable and the number of local maxima of the field above a threshold u in an R×R box in R2 decays like exp(−βu^2), for some fixed β > 0. As an immediate consequence, when the level u is a function of R such that u(R) →∞ and u(R)/√log R →0 as R →∞, we have a quantitative central limit theorem for the number of high local maxima. The proof is based on the Chen-Stein method for quantitative Poisson approximation [CX04]. The basic idea is that, for any point process η , if η and its Palm version are “close” in some (pseudo)metric, then η is “close” to a Poisson point process. Here, we produce a close coupling of a (non-degenerate) stationary smooth field and its Palm version. Then, we study the difference in the number of critical points using smooth flows of critical points. We believe the same method/argument works for any non-degenerate stationary Gaussian field with sufficient correlation decay and in any dimension.This thesis’s critical point structure results are joint with Dmitry Belyaev
Function meets circularity: Metal-ionomer crosslinks toughen and recycle CO2-derived polymers - data
No full textSupporting data for publication "Function Meets Circularity: Metal-Ionomer Crosslinks Toughen and Recycle CO2-Derived Polymers"
The complexity of explicit construction problems
Get PDFExplicit constructions of pseudorandom objects play a central role in theoretical computer science. Unfortunately, for many properties of interest, while a randomly chosen object would satisfy them with high probability, deterministically constructing such objects remains notoriously challenging. For example, Erdős famously showed that a random graph is almost always a Ramsey graph, yet explicitly constructing Ramsey graphs has been a long-standing open problem.
Recently, several explicit construction problems have seen progress through complexity-theoretic ideas. Unlike previous ad-hoc methods tailored to individual cases, these complexity-theoretic approaches offer systematic solutions applicable to broad families of explicit construction problems. Central to this line of research is the Range Avoidance problem (Avoid), which encapsulates a wide array of explicit construction problems: Deterministically solving Avoid would simultaneously resolve many open questions in explicit construction. What are the strengths and limitations of these complexity-theoretic techniques, and how can understanding the complexity of Avoid illuminate these questions?
This thesis examines the complexity of explicit construction problems, with a particular focus on the Range Avoidance problem. We uncover a strong connection between explicit construction problems and fundamental questions in complexity theory, making progress in both domains. Specifically, our contributions include:
* Algorithmic Method for Range Avoidance: We generalise Williams's Algorithmic Method (2011)—originally developed for proving circuit lower bounds—to solve the Range Avoidance problem. Our results demonstrate that this method applies not only to circuit lower bounds but also to explicit construction problems in general. Consequently, we derive new complexity lower bounds and develop novel algorithms for special cases of Avoid. Building on these techniques, we further show that a slight extension of the Algorithmic Method fully characterises, i.e., is both necessary and sufficient for, proving circuit lower bounds for E^NP.
* Unconditional constructions: We present new unconditional results in explicit constructions. In particular, we devise an infinitely-often pseudodeterministic polynomial-time algorithm for finding prime numbers. We also introduce a new algorithm for solving the Range Avoidance problem. The latter result yields near-optimal circuit lower bounds for the complexity classes Σ2E and S2E/1, resolving a 40-year-old open question from Kannan (1982). These results are obtained through a novel "iterative win-win" method, which is likely to have broader applications in complexity theory.
* Additional results: In addition, we study the related Heavy Range Avoidance problem, uncovering its connections to uniform lower bounds and derandomisation. We also present new hardness results for the Range Avoidance problem under Rudich’s demi-bits conjectures, which have implications in proof complexity as well
A thirty-four-marker spectral flow cytometry panel for deep immunophenotyping to characterize activation, cytotoxic and metabolic features of unconventional and conventional T cells in human peripheral blood
No full textThis data is part of a manuscript describing the development and validation of a spectral flow cytometry panel