Biology subjects

Luthra, S.

Publications and source records attributed to Luthra, S..

3 recordsLinked to original sources

bioRxiv2025

Beyond the Hype: The Complexity of Automated Cell Type Annotations with GPT-4

Recent research has shown the impressive capability of large language models like GPT-4 in various downstream tasks in single-cell data analysis. Among these tasks, cell type annotation remains particularly challenging, with researchers exploring various methods to improve accuracy and efficiency. While recent studies on GPT-like models have demonstrated annotation performance comparable to manual annotations, a significant gap remains in understanding their limitations and generalizability. In this work, we compare and evaluate the annotation performance of the GPT-4 model against traditional methods on nine randomly selected public single-cell RNA seq datasets from cellxgene, covering diverse tissue types. Our evaluation highlights the complexity of annotating cell types in single-cell data, revealing key differences between automated and manual approaches. We found specific cases where GPT-4 underperforms, demonstrating its limitations in certain contexts. We further introduce an automated approach to incorporate literature search using a RAG approach which enhances and outperforms GPT-4 cell type annotation when compared to traditional methods. We also introduce metrics based on taxonomic distance in the ontology tree to evaluate the granularity of the cell type annotations. To support future research, we also release an open-source Python package 1 that enables fully automated cell-type annotation of single-cell data using GPT-4 alongside other methods. The pipeline can take paper as an input and do cell type annotations on its own.

bioinformatics↗

bioRxiv2024

Systematic changes in neural selectivity reflect the acquired salience of category-diagnostic dimensions

Humans and other animals develop remarkable perceptual and cognitive specializations for identifying, differentiating, and acting on classes of ecologically important signals. This expertise is flexible enough to support diverse perceptual judgments: a voice, for example, simultaneously conveys what a talker says, as well as myriad cues about her identity and state. Expert perception across complex signals thus involves discovering and learning regularities that best inform diverse perceptual judgments, as well as weighting this information flexibly as task demands change. Here, we test whether this flexibility may involve endogenous attentional gain. We use two prospective auditory category learning tasks to relate a complex, entirely novel soundscape to four classes of "alien identity" and two classes of "alien size." Identity, but not size, categorization requires discovery and learning of patterned acoustic input situated in one of two simultaneous, non-overlapping frequency bands. This allows us to capitalize on the coarsely segregated frequency-band-specific channels tiling auditory cortex, using fMRI to ask whether category-relevant perceptual information present in one frequency band is prioritized relative to simultaneous, uninformative information in the other frequency band. Among participants expert at alien identity categorization, we observe prioritization of the identity-diagnostic frequency band that persists even when the diagnostic information becomes irrelevant in the size categorization task. Tellingly, the neural selectivity evoked implicitly in the identity categorization task aligns with that in an independent task, where activation is driven by explicit and sustained selective attention to pure tones in one or the other frequency band. Additionally, the learning trajectories taken to achieve expert-level categorization leave fingerprints on the patterns of neural activity associated with the diagnostic dimension. In all, this indicates that acquiring categories can drive the emergence of acquired attentional gain to category-diagnostic input dimensions.

neuroscience↗

bioRxiv2024

Distributional learning drives statistical deafening

Humans and other animals use information about how likely it is for something to happen. The absolute and relative probability of an event influences a remarkable breadth of behaviors, from foraging for food to comprehending linguistic constructions -- even when these probabilities are learned implicitly. It is less clear how, and under what circumstances, statistical learning of simple probabilities might drive changes in perception and cognition. Here, across a series of 29 experiments, we probe listeners sensitivity to task-irrelevant changes in the probability distribution of tones acoustic frequency across tone-in-noise detection and tone duration decisions. We observe that the task-irrelevant frequency distribution influences the ability to detect a sound and the speed with which perceptual decisions about its duration are made. The shape of the probability distribution, its range, and a tones relative position within that range impact observed patterns of suppression and enhancement of tone detection and decision making. Perceptual decisions are also modulated by a newly discovered perceptual bias, with lower frequencies in the distribution more often and more rapidly perceived as longer, and higher frequencies as shorter. Perception is sensitive to rapid distribution changes, but distributional learning from previous probability distributions also carries over. In fact, massed exposure to a single point along the dimension results in seemingly maladaptive loss of sensitivity - occurring entirely in the absence of feedback or reward - along a range of subsequently encountered frequencies. This points to a gain mechanism that suppresses sensitivity to regions along a perceptual dimension that are less likely to be encountered. Significance StatementOrganisms as diverse as honeybees and humans pick up on probabilities in the world around them. People implicitly learn the likelihood of a color, price range, or even syntactic structure. How does statistical learning affect how we detect events and make decisions, especially when probabilities are completely irrelevant to the task at hand, and can change without warning? We find that people learn and track changes in perceptual probabilities irrelevant to a task and that this learning drives dynamic shifts in perception characterized by graded effects of enhancement - and primarily - suppression across acoustic frequency. This can result in a remarkably long- lived diminishment of perceptual sensitivity that seems maladaptive but may instead reflect use of likelihood to guide and sharpen perception.

neuroscience↗