Biology subjects

Nemati, S.

Publications and source records attributed to Nemati, S..

3 recordsLinked to original sources

bioRxiv2026

Head-level lesion-symptom mapping of picture naming in vision-language models

Researchers in artificial intelligence increasingly intervene on language models to study how their functions are organized, silencing weights and attention components in ways reminiscent of the brain lesions long used to map language in post-stroke aphasia. Under this program of mechanistic interpretability, a common move is to ablate an attention head and read the resulting drop in a behavior as evidence that the head implements it. How much that move can establish about where a behavior is computed remains unclear, because a head whose removal disrupts a behavior is necessary for it but need not be the site where it is computed. We examined this question for the picture naming task, taking the loss of naming (anomia) that defines aphasia as the behavior of interest. Across six vision-language models, spanning three language backbones and a range of parameter scales, we ablated each attention head in turn during single-word picture naming and measured accuracy before and after. The degree of localization varied widely across models. In LLaVA-1.6-Vicuna-13B, a single early head (layer 0, head 20) was necessary: removing it alone reduced naming accuracy from 0.99 to 0.006. The same head was not sufficient, because retaining it while ablating the other 1{,}599 heads also produced 0\% accuracy. Two Mistral-backbone models (LLaVA-Mistral-7B and Idefics2-8B) had no critical head. An early-layer dependence was present in every model but varied in strength, whereas dependence on any single head ranged from dominant to absent. Within Qwen2.5-VL, a dominant head was present in the 7B model but not the 3B model, indicating that this concentration emerged with scale rather than being fixed across a model family, and it was not explained by attention type. These results show that ablating the head whose removal disrupts naming does not establish that the head computes the behavior, that the result generalizes across models, or that the behavior localizes to a head at all.

neuroscience↗

bioRxiv2024

TXNIP mediates LAT1/SLC7A5 endocytosis to reduce amino acid uptake in cells entering quiescence

Entry and exit from cellular quiescence require dynamic adjustments in nutrient acquisition, yet the mechanisms by which quiescent cells downregulate amino acid (AA) transport remain poorly understood. Here, we demonstrate that cells entering quiescence select plasma membrane-resident AA transporters for endocytosis and lysosomal degradation, to match AA uptake with reduced translation. We identify the -arrestin TXNIP as a key regulator of AA uptake during quiescence, since it mediates the endocytosis of the SLC7A5-SLC3A2 (LAT1-4F2hc) transporter complex in response to reduced AKT signaling. Mechanistically, TXNIP interacts with HECT-type ubiquitin ligases to facilitate transporter ubiquitination. Loss of TXNIP disrupts this regulation, resulting in dysregulated AA uptake, sustained mTORC1 signaling, and accelerated quiescence exit. A novel TXNIP loss-of-function mutation in a patient with severe metabolic disease further supports its role in nutrient homeostasis and human health. These findings highlight TXNIPs role in controlling SLC7A5-SLC3A2 mediated AA acquisition with implications for quiescence biology and disease.

cell biology↗

bioRxiv2021

Density Fluctuations Yield Distinct Growth and Fitness Effects in Single Bacteria

Single-cells grow by increasing their biomass and size. Here, we report that while mass and size accumulation rates of single Escherichia coli cells are exponential, their density fluctuates during growth. As such, the rates of mass and size accumulation of a single-cell are generally not the same, but rather cells differentiate into increasing one rate with respect to the other. This differentiation yields a previously unknown density homeostasis mechanism, which we support mathematically. Further, growth differentiation challenges ongoing efforts to predict single-cell reproduction rates (or fitness-levels), through the accumulation rates of size or mass. In contrast, we observe that density fluctuations can predict fitness, with only high fitness individuals existing in the high density fluctuation regime. We detail our imaging approach and the invisible microfluidic arrays that critically enabled increased precision and throughput. Biochemical production, infections, and natural communities start from few, growing, cells, thus, underscoring the significance of density-fluctuations when considering non-genetic variability.

biophysics↗