arXiv subjects

Aaron Sim

Publications and source records attributed to Aaron Sim.

11 recordsLinked to original sources

arXiv2025

On fine-tuning Boltz-2 for protein-protein affinity prediction

Accurate prediction of protein-protein binding affinity is vital for understanding molecular interactions and designing therapeutics. We adapt Boltz-2, a state-of-the-art structure-based protein-ligand affinity predictor, for protein-protein affinity regression and evaluate it on two datasets, TCR3d and PPB-affinity. Despite high structural accuracy, Boltz-2-PPI underperforms relative to sequence-based alternatives in both small- and larger-scale data regimes. Combining embeddings from Boltz-2-PPI with sequence-based embeddings yields complementary improvements, particularly for weaker sequence models, suggesting different signals are learned by sequence- and structure-based models. Our results echo known biases associated with training with structural data and suggest that current structure-based representations are not primed for performant affinity prediction.

cs.LG↗

arXiv2023

Proxy-based Zero-Shot Entity Linking by Effective Candidate Retrieval

A recent advancement in the domain of biomedical Entity Linking is the development of powerful two-stage algorithms, an initial candidate retrieval stage that generates a shortlist of entities for each mention, followed by a candidate ranking stage. However, the effectiveness of both stages are inextricably dependent on computationally expensive components. Specifically, in candidate retrieval via dense representation retrieval it is important to have hard negative samples, which require repeated forward passes and nearest neighbour searches across the entire entity label set throughout training. In this work, we show that pairing a proxy-based metric learning loss with an adversarial regularizer provides an efficient alternative to hard negative sampling in the candidate retrieval stage. In particular, we show competitive performance on the recall@1 metric, thereby providing the option to leave out the expensive candidate ranking step. Finally, we demonstrate how the model can be used in a zero-shot setting to discover out of knowledge base biomedical entities.

cs.LG↗

arXiv2022

Pseudo-Riemannian Embedding Models for Multi-Relational Graph Representations

In this paper we generalize single-relation pseudo-Riemannian graph embedding models to multi-relational networks, and show that the typical approach of encoding relations as manifold transformations translates from the Riemannian to the pseudo-Riemannian case. In addition we construct a view of relations as separate spacetime submanifolds of multi-time manifolds, and consider an interpolation between a pseudo-Riemannian embedding model and its Wick-rotated Riemannian counterpart. We validate these extensions in the task of link prediction, focusing on flat Lorentzian manifolds, and demonstrate their use in both knowledge graph completion and knowledge discovery in a biological domain.

cs.SI↗

arXiv2022

Phenotyping with Positive Unlabelled Learning for Genome-Wide Association Studies

Identifying phenotypes plays an important role in furthering our understanding of disease biology through practical applications within healthcare and the life sciences. The challenge of dealing with the complexities and noise within electronic health records (EHRs) has motivated applications of machine learning in phenotypic discovery. While recent research has focused on finding predictive subtypes for clinical decision support, here we instead focus on the noise that results in phenotypic misclassification, which can reduce a phenotypes ability to detect associations in genome-wide association studies (GWAS). We show that by combining anchor learning and transformer architectures into our proposed model, AnchorBERT, we are able to detect genomic associations only previously found in large consortium studies with 5$\times$ more cases. When reducing the number of controls available by 50\%, we find our model is able to maintain 40\% more significant genomic associations from the GWAS catalog compared to standard phenotype definitions. \keywords{Phenotyping \and Machine Learning \and Semi-Supervised \and Genetic Association Studies \and Biological Discovery}

stat.AP↗

arXiv2021

Directed Graph Embeddings in Pseudo-Riemannian Manifolds

The inductive biases of graph representation learning algorithms are often encoded in the background geometry of their embedding space. In this paper, we show that general directed graphs can be effectively represented by an embedding model that combines three components: a pseudo-Riemannian metric structure, a non-trivial global topology, and a unique likelihood function that explicitly incorporates a preferred direction in embedding space. We demonstrate the representational capabilities of this method by applying it to the task of link prediction on a series of synthetic and real directed graphs from natural language applications and biology. In particular, we show that low-dimensional cylindrical Minkowski and anti-de Sitter spacetimes can produce equal or better graph representations than curved Riemannian manifolds of higher dimensions.

stat.ML↗

arXiv2021

Contrastive Mixture of Posteriors for Counterfactual Inference, Data Integration and Fairness

Learning meaningful representations of data that can address challenges such as batch effect correction and counterfactual inference is a central problem in many domains including computational biology. Adopting a Conditional VAE framework, we show that marginal independence between the representation and a condition variable plays a key role in both of these challenges. We propose the Contrastive Mixture of Posteriors (CoMP) method that uses a novel misalignment penalty defined in terms of mixtures of the variational posteriors to enforce this independence in latent space. We show that CoMP has attractive theoretical properties compared to previous approaches, and we prove counterfactual identifiability of CoMP under additional assumptions. We demonstrate state-of-the-art performance on a set of challenging tasks including aligning human tumour samples with cancer cell-lines, predicting transcriptome-level perturbation responses, and batch correction on single-cell RNA sequencing data. We also find parallels to fair representation learning and demonstrate that CoMP is competitive on a common task in the field.

stat.ML↗

arXiv2018

Interpretable Graph Convolutional Neural Networks for Inference on Noisy Knowledge Graphs

In this work, we provide a new formulation for Graph Convolutional Neural Networks (GCNNs) for link prediction on graph data that addresses common challenges for biomedical knowledge graphs (KGs). We introduce a regularized attention mechanism to GCNNs that not only improves performance on clean datasets, but also favorably accommodates noise in KGs, a pervasive issue in real-world applications. Further, we explore new visualization methods for interpretable modelling and to illustrate how the learned representation can be exploited to automate dataset denoising. The results are demonstrated on a synthetic dataset, the common benchmark dataset FB15k-237, and a large biomedical knowledge graph derived from a combination of noisy and clean data sources. Using these improvements, we visualize a learned model's representation of the disease cystic fibrosis and demonstrate how to interrogate a neural network to show the potential of PPARG as a candidate therapeutic target for rheumatoid arthritis.

cs.LG↗

arXiv2015

Great cities look small

Great cities connect people; failed cities isolate people. Despite the fundamental importance of physical, face-to-face social-ties in the functioning of cities, these connectivity networks are not explicitly observed in their entirety. Attempts at estimating them often rely on unrealistic over-simplifications such as the assumption of spatial homogeneity. Here we propose a mathematical model of human interactions in terms of a local strategy of maximising the number of beneficial connections attainable under the constraint of limited individual travelling-time budgets. By incorporating census and openly-available online multi-modal transport data, we are able to characterise the connectivity of geometrically and topologically complex cities. Beyond providing a candidate measure of greatness, this model allows one to quantify and assess the impact of transport developments, population growth, and other infrastructure and demographic changes on a city. Supported by validations of GDP and HIV infection rates across United States metropolitan areas, we illustrate the effect of changes in local and city-wide connectivities by considering the economic impact of two contemporary inter- and intra-city transport developments in the United Kingdom: High Speed Rail 2 and London Crossrail. This derivation of the model suggests that the scaling of different urban indicators with population size has an explicitly mechanistic origin.

physics.soc-ph↗

arXiv2013

Random Forests on Distance Matrices for Imaging Genetics Studies

We propose a non-parametric regression methodology, Random Forests on Distance Matrices (RFDM), for detecting genetic variants associated to quantitative phenotypes representing the human brain's structure or function, and obtained using neuroimaging techniques. RFDM, which is an extension of decision forests, requires a distance matrix as response that encodes all pair-wise phenotypic distances in the random sample. We discuss ways to learn such distances directly from the data using manifold learning techniques, and how to define such distances when the phenotypes are non-vectorial objects such as brain connectivity networks. We also describe an extension of RFDM to detect espistatic effects while keeping the computational complexity low. Extensive simulation results and an application to an imaging genetics study of Alzheimer's Disease are presented and discussed.

stat.ML↗

arXiv2012

Information Geometry and Sequential Monte Carlo

This paper explores the application of methods from information geometry to the sequential Monte Carlo (SMC) sampler. In particular the Riemannian manifold Metropolis-adjusted Langevin algorithm (mMALA) is adapted for the transition kernels in SMC. Similar to its function in Markov chain Monte Carlo methods, the mMALA is a fully adaptable kernel which allows for efficient sampling of high-dimensional and highly correlated parameter spaces. We set up the theoretical framework for its use in SMC with a focus on the application to the problem of sequential Bayesian inference for dynamical systems as modelled by sets of ordinary differential equations. In addition, we argue that defining the sequence of distributions on geodesics optimises the effective sample sizes in the SMC run. We illustrate the application of the methodology by inferring the parameters of simulated Lotka-Volterra and Fitzhugh-Nagumo models. In particular we demonstrate that compared to employing a standard adaptive random walk kernel, the SMC sampler with an information geometric kernel design attains a higher level of statistical robustness in the inferred parameters of the dynamical systems.

stat.ME↗

arXiv2009

E7(7) formulation of N=2 backgrounds

In this paper we reformulate N=2 supergravity backgrounds arising in type II string theory in terms of quantities transforming under the U-duality group E7(7). In particular we combine the Ramond--Ramond scalar degrees of freedom together with the O(6,6) pure spinors which govern the Neveu-Schwarz sector by considering an extended version of generalised geometry. We give E7(7)-invariant expressions for the Kahler and hyperkahler potentials describing the moduli space of vector and hypermultiplets, demonstrating that both correspond to standard E7(7) coset spaces. We also find E7(7) expressions for the Killing prepotentials defining the scalar potential, and discuss the equations governing N=1 vacua in this formalism.

hep-th↗