arXiv subjects

Sohini Ramachandran

Publications and source records attributed to Sohini Ramachandran.

4 recordsLinked to original sources

arXiv2025

Principles and Policy Recommendations for Comprehensive Genetic Data Governance

Genetic data collection has become ubiquitous, producing genetic information about health, ancestry, and social traits. However, unregulated use, especially amid evolving scientific understanding, poses serious privacy and discrimination risks. These risks are intensified by advancing AI, particularly multi-modal systems integrating genetic, clinical, behavioral, and environmental data. In this work, we organize the uses of genetic data along four distinct "pillars", and develop a risk assessment framework that identifies key values any governance system must preserve. In doing so, we draw on current privacy scholarship concerning contextual integrity, data relationality, and the Belmont principle. We apply the framework to four real-world case studies and identify critical gaps in existing regulatory frameworks and specific threats to privacy and personal liberties, particularly through genetic discrimination. Finally, we offer three policy recommendations for genetic data governance that safeguard individual rights in today's under-regulated ecosystem of large-scale genetic data collection and usage.

cs.CY↗

arXiv2022

The Drift of #MyBodyMyChoice Discourse on Twitter

#MyBodyMyChoice is a well-known hashtag originally created to advocate for women's rights, often used in discourse about abortion and bodily autonomy. The Covid-19 outbreak prompted governments to take containment measures such as vaccination campaigns and mask mandates. Population groups opposed to such measures started to use the slogan "My Body My Choice" to claim their bodily autonomy. In this paper, we investigate whether the discourse around the hashtag #MyBodyMyChoice on Twitter changed its usage after the Covid-19 outbreak. We observe that the conversation around the hashtag changed in two ways. First, semantically, the hashtag #MyBodyMyChoice drifted towards conversations around Covid-19, especially in messages opposed to containment measures. Second, while before the pandemic users used to share content produced by experts and authorities, after Covid-19 the users' attention has shifted towards individuals.

cs.CY↗

arXiv2022

Human genetic admixture through the lens of population genomics

Over the last fifty years, geneticists have made great strides in understanding how our species' evolutionary history gave rise to current patterns of human genetic diversity classically summarized by Lewontin in his 1972 paper, 'The Apportionment of Human Diversity'. One evolutionary process that requires special attention in both population genetics and statistical genetics is admixture: gene flow between two or more previously separated source populations to form a new admixed population. The admixture process introduces unique patterns of genetic variation within and between populations, which in turn influences the inference of demographic histories, identification genetic targets of selection, and prediction of phenotypes. In this review, we highlight recent studies and methodological advances that have leveraged genomic signatures of admixture to gain insights into human history, natural selection, and complex trait architecture. We also outline some challenges for admixture population genetics, including limitations of applying methods designed for single-ancestry populations to the study of admixed populations.

q-bio.PE↗

arXiv2020

Generalizing Variational Autoencoders with Hierarchical Empirical Bayes

Variational Autoencoders (VAEs) have experienced recent success as data-generating models by using simple architectures that do not require significant fine-tuning of hyperparameters. However, VAEs are known to suffer from over-regularization which can lead to failure to escape local maxima. This phenomenon, known as posterior collapse, prevents learning a meaningful latent encoding of the data. Recent methods have mitigated this issue by deterministically moment-matching an aggregated posterior distribution to an aggregate prior. However, abandoning a probabilistic framework (and thus relying on point estimates) can both lead to a discontinuous latent space and generate unrealistic samples. Here we present Hierarchical Empirical Bayes Autoencoder (HEBAE), a computationally stable framework for probabilistic generative models. Our key contributions are two-fold. First, we make gains by placing a hierarchical prior over the encoding distribution, enabling us to adaptively balance the trade-off between minimizing the reconstruction loss function and avoiding over-regularization. Second, we show that assuming a general dependency structure between variables in the latent space produces better convergence onto the mean-field assumption for improved posterior inference. Overall, HEBAE is more robust to a wide-range of hyperparameter initializations than an analogous VAE. Using data from MNIST and CelebA, we illustrate the ability of HEBAE to generate higher quality samples based on FID score than existing autoencoder-based approaches.

stat.ML↗