Biology subjects

Yildirim, I.

Publications and source records attributed to Yildirim, I..

8 recordsLinked to original sources

bioRxiv2026

Hypothesis-Driven Identification of Neural Algorithms With Dynamical Structure-Preserving Manifolds.

Neuroscience aims to uncover the algorithms by which the brain builds and manipulates rich internal representations of objects, agents, and places. Existing reverse-engineering frameworks remain ineffective for identifying these algorithms. Here, we present Dynamical Structure-Preserving Manifolds (dSPMs)---a novel multilevel reverse-engineering framework that programs, without training, recurrent neural networks to enable hypothesis-driven identification of the brain's algorithms that generate symbolic, structure-preserving representations of the world around. We apply dSPM to identify an algorithm for physical scene prediction in the dorsomedial frontal cortex (DMFC) of macaques intercepting a ball on a Pong-like virtual gameboard. dSPM better explained DMFC than alternatives, made predictions confirmed in neural data, and revealed an algorithm of DMFC that we term 'manifold-based simulation': perceived initial conditions rapidly collapse neural activity onto a low-dimensional manifold whose geometry encodes a physics-based representation of the scene, supporting far-out predictions without traditional next-step simulation. dSPM supports effective exploration of the computational foundations of biological intelligence.

neuroscience↗

bioRxiv2024

The invasion phenotypes of glioblastoma depend on plastic and reprogrammable cell states

Glioblastoma (GBM), the most common primary brain cancer in adults, is characterized by rapid local invasion along diverse routes, such as infiltration of white matter tracts and penetration of perivascular spaces. We investigate the hypothesis that GBM invasion routes correlate with the transcriptional states of individual cells and identify regulators of route-specific invasion. Utilizing patient-derived GBM xenograft models, we integrate single-cell transcriptomics and spatial proteomics, revealing that mesenchymal and oligodendrocyte progenitor-like GBM cells migrate perivascularly, while neural progenitor and astrocyte-like GBM cells invade diffusely. Computational reconstruction identifies ANXA1 as a perivascular invasion driver and lineage-restricted transcription factors RFX4 and HOPX as drivers of diffuse invasion, predictive of patient survival. Genetic ablation of these genes alters invasion phenotypes and extends survival in xenografted mice, clarifying the role of cell states in GBM invasion, and highlighting potential therapeutic targets for selective invasion route targeting in GBM patients.

cancer biology↗

bioRxiv2023

Temporal segmentation and 'look ahead' simulation: Physical events structure visual perception of intuitive physics

How we perceive the physical world is not only organized in terms of objects, but also structured in time as sequences of events. This is especially evident in intuitive physics, with temporally bounded dynamics such as falling, occlusion, and bouncing demarcating the continuous flow of sensory inputs. While the spatial structure and attentional consequences of physical objects have been well-studied, much less is known about the temporal structure and attentional consequences of physical events in visual perception. Previous work has recognized physical events as units in the mind, and used pre-segmented object interactions to explore physical representations. However, these studies did not address whether and how perception imposes the kind of temporal structure that carves these physical events to begin with, and the attentional consequences of such segmentation during intuitive physics. Here, we use performance-based tasks to address this gap. In Experiment 1, we find that perception not only spontaneously separates visual input in time into physical events, but also, this segmentation occurs in a nonlinear manner within a few hundred milliseconds at the moment of the event boundary. In Experiment 2, we find that event representations, once formed, use coarse look ahead simulations to selectively prioritize those objects that are predictively part of the unfolding dynamics. This rich temporal and predictive structure of physical events, formed during vision, should inform models of intuitive physics.

neuroscience↗

bioRxiv2022

Evaluating Quantitative and Functional MRI As Potential Techniques to Identify the Subdivisions in the Human Lateral Geniculate Nucleus

Segmenting the magnocellular (M) and parvocellular (P) divisions of the human lateral geniculate nucleus (LGN) has been challenging yet remains an important goal because the LGN is the only place in the brain where these two information streams are spatially disjoint and can be studied independently. Previous research used the amplitude of responses to different types of stimuli to separate M and P regions (Denison et al., 2014; Zhang et al., 2015). However, this method is confounded because the hilum region of the LGN exhibits greater response amplitudes to all stimuli and can be mistaken for the M subdivision (DeSimone & Schneider, 2019). Therefore, we have employed two independent methodologies that do not rely upon the functional response properties of the M and P neurons to segment the M and P regions: 1) structural quantitative MRI (qMRI) at 3T to measure the T1 relaxation time, and 2) monocular and dichoptic functional MRI (fMRI) procedures to measure eye-specific responses. Our qMRI results agreed with the anatomical expectations, identifying M regions on the ventromedial surface of the LGN. The monocular fMRI procedure was better than the dichoptic condition to identify the eye-dominance signals. Both procedures revealed significant right eye bias, and neither could reliably identify the first M layer of the LGN. These findings indicated that the qMRI methods are promising whereas the functional identification of contralateral layers requires further refinement. HighlightsO_LIT1 parameter in qMRI segregates M and P regions of LGN in individual subjects at 3T. C_LIO_LIEye-specific voxels in LGN respond more strongly to monocular than dichoptic viewing. C_LIO_LIClusters of eye-specific regions but not layers can be separated at 1.5 mm resolution. C_LI

neuroscience↗

bioRxiv2021

No evidence of fetal defects or anti-syncytin-1 antibody induction following COVID-19 mRNA vaccination

The impact of coronavirus disease 2019 (COVID-19) mRNA vaccination on pregnancy and fertility has become a major topic of public interest. We investigated two of the most widely propagated claims to determine 1) whether COVID-19 mRNA vaccination of mice during early pregnancy is associated with an increased incidence of birth defects or growth abnormalities, and 2) whether COVID-19 mRNA-vaccinated human volunteers exhibit elevated levels of antibodies to the human placental protein syncytin-1. Using a mouse model, we found that intramuscular COVID-19 mRNA vaccination during early pregnancy at gestational age E7.5 did not lead to differences in fetal size by crown-rump length or weight at term, nor did we observe any gross birth defects. In contrast, injection of the TLR3 agonist and double-stranded RNA mimic polyinosinic-polycytidylic acid, or poly(I:C), impacted growth in utero leading to reduced fetal size. No overt maternal illness following either vaccination or poly(I:C) exposure was observed. We also found that term fetuses from vaccinated murine pregnancies exhibit high circulating levels of anti-Spike and anti-RBD antibodies to SARS-CoV-2 consistent with maternal antibody status, indicating transplacental transfer. Finally, we did not detect increased levels of circulating anti-syncytin-1 antibodies in a cohort of COVID-19 vaccinated adults compared to unvaccinated adults by ELISA. Our findings contradict popular claims associating COVID-19 mRNA vaccination with infertility and adverse neonatal outcomes.

immunology↗

bioRxiv2021

Perception of soft materials relies on physics-based object representations: Behavioral and computational evidence

When encountering objects, we readily perceive not only low-level properties (e.g., color and orientation), but also seemingly higher-level ones - including aspects of physics (e.g., mass). Perhaps nowhere is this contrast more salient than in the perception of soft materials such as cloths: the dynamics of these objects (including how their three-dimensional forms vary) are determined by their physical properties such as stiffness, elasticity, and mass. Here we hypothesize that the perception of cloths and their physical properties must involve not only image statistics, but also abstract object representations that incorporate "intuitive physics". We provide behavioral and computational evidence for this hypothesis. We find that humans can visually match the stiffness of cloths with unfamiliar textures from the way they undergo natural transformations (e.g. flapping in the wind) across different scenarios. A computational model that casts cloth perception as mental physics simulation explains important aspects of this behavior.

neuroscience↗

bioRxiv2020

A small molecule that binds an RNA repeat expansion stimulates its decay via the exosome complex

We describe the design of a small molecule that binds the structure of a r(CUG) repeat expansion [r(CUG)exp] and reverses molecular defects in two diseases mediated by the RNA - myotonic dystrophy type 1 (DM1) and Fuchs endothelial corneal dystrophy (FECD). Thus, a single structure-specific ligand has potential therapeutic benefit for multiple diseases, in contrast to oligonucleotide-based modalities that are customized for each disease by nature of targeting the gene that harbors the repeat. Indeed, the small molecule binds the target with nanomolar affinity and >100-fold specificity vs. many other RNAs and DNA. Interestingly, the compounds downstream effects are different in the two diseases, owing to the location of the repeat expansion. In DM1, r(CUG)exp is harbored in the 3 untranslated region (UTR) of and mRNA, and the compound has no effect on the RNAs abundance. In FECD, however, r(CUG)exp is located in an intron, and the small molecule, by binding the repeat expansion, facilitates excision of the intron, which is then degraded by the exosome complex exonuclease, hRRP6. Thus, structure-specific, RNA-targeting small molecules can act disease-specifically to affect biology, either by disabling its gain-of-function mechanism (DM1) or by stimulating quality control pathways to rid a disease-affected cell of a toxic RNA (FECD). Significance statementMany different diseases are caused by toxic structured RNAs. Herein, we designed a lead small molecule that binds a toxic structure and rescues disease biology. We show that a structure-specific small molecule can improve disease-associated defects in two diseases that share the common toxic RNA structure. In one disease, the toxic structure is harbored in an intron and causes its retention. The compound facilitates processing of a retained intron, enabling the disease-affected cell to remove the toxic RNA.

cell biology↗

bioRxiv2018

Efficient inverse graphics in biological face processing

Vision must not only recognize and localize objects, but perform richer inferences about the underlying causes in the world that give rise to sensory data. How the brain performs these inferences remains unknown: Theoretical proposals based on inverting generative models (or "analysis-by-synthesis") have a long history but their mechanistic implementations have typically been too slow to support online perception, and their mapping to neural circuits is unclear. Here we present a neurally plausible model for efficiently inverting generative models of images and test it as an account of one high-level visual capacity, the perception of faces. The model is based on a deep neural network that learns to invert a three-dimensional (3D) face graphics program in a single fast feedforward pass. It explains both human behavioral data and multiple levels of neural processing in non-human primates, as well as a classic illusion, the "hollow face" effect. The model fits qualitatively better than state-of-the-art computer vision models, and suggests an interpretable reverse-engineering account of how images are transformed into percepts in the ventral stream.

neuroscience↗