Biology subjects

Sulam, J.

Publications and source records attributed to Sulam, J..

8 recordsLinked to original sources

bioRxiv2026

Real-time AI integration for MR to detect artifacts and guide pulse sequence adaptations

This work presents a first-of-its-kind artificial intelligence (AI-)integrated MR pulse sequence that detects out-of-voxel (OOV) artifacts in real-time (within-TR) and responds prospectively by updating the crusher gradient scheme. Per Excitation Real-time Execution & Guided Responses with Integrated Neural-network Evaluation (PEREGRINE), allows for deployment of deep learning models and pulse sequence updates. In this study, PEREGRINE operated a time-domain (TD) and frequency-domain (FD) convolutional autoencoder that detect OOV artifacts. Scans without (AI-off) and with (AI-on) updates were collected from the medial prefrontal cortex of healthy volunteers using a MEGA-edited MRS experiment. The degree of OOV contamination (OOV Score) was quantified per transient based upon the prevalence of OOV signals in the TD and FD data. OOV Scores above a user-defined threshold triggered an update of the crusher gradient scheme, iterating through 48 permutations (6 axis transpositions x 8 polarity flips). Within each 2-second TR, PEREGRINE successfully provided single-transient OOV Scores and updated gradients accordingly. No difference was observed between the OOV Scores from the full ("Full" condition) AI-on and AI-off sessions due to the AI-on scan cycling over better and worse gradient permutations relative to the AI-off scan. However, the AI-on scan had significantly lower OOV Scores than the AI-off scan when selecting the transients where PEREGRINE persisted ("Dwell" condition) on a given gradient permutation. Ultimately, Fit Quality Number (FQN) from linear combination modeling improved significantly for the AI-on compared to the AI-off scan. PEREGRINE enabled an adaptive and AI-integrated sequence allowing for real-time evaluation and response to OOV artifacts, identifying gradient modifications that produced less OOV contamination. Graphical AbstractO_ST_ABSGraphical Abstract SummaryC_ST_ABSPEREGRINE, Per Excitation Real-time Execution & Guided Responses with Integrated Neural-network Evaluation, allows for pulse sequence updates using on-scanner deployed neural networks. To demonstrate the utility, PEREGRINE was used to operate two neural networks (time/frequency domain), trained to detect out-of-voxel artifacts, during a MEGA-edited MRS experiment in the prefrontal cortex. When artifacts were detected, PEREGRINE updated the crusher gradients within the same TR. Real-time, adaptive, and AI-driven MR will provide a long-awaited solution for combatting artifacts and poor data quality. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=167 SRC="FIGDIR/small/722724v2_ufig1.gif" ALT="Figure 1"> View larger version (38K): org.highwire.dtl.DTLVardef@cc01e0org.highwire.dtl.DTLVardef@18f4083org.highwire.dtl.DTLVardef@1d48c8corg.highwire.dtl.DTLVardef@1574ba7_HPS_FORMAT_FIGEXP M_FIG Graphical Abstract Figure C_FIG

neuroscience↗

bioRxiv2024

Brain-wide mapping of oligodendrocyte organization and oligodendrogenesis across the murine lifespan

Insulating sheaths of myelin accelerate neuronal signaling in complex networks of the mammalian brain. In the CNS, myelin sheaths are exclusively produced by oligodendrocytes, which continue to be generated throughout life to change patterns of myelination. However, a brain-wide analysis of oligodendrocyte dynamics across the lifespan has not been performed. We developed a rapid, robust cellular mapping pipeline involving tissue clearing, lightsheet microscopy, atlas alignment, and automated segmentation to define the location of all oligodendrocytes in the mouse brain. This analysis demonstrated the remarkable consistency of oligodendrocyte patterns between hemispheres, individuals, and sexes, and established that oligodendrocyte maps estimate myelin coverage. We trained a vision transformer to identify newly generated oligodendrocytes from millions of mature cells, highlighting age- and region-specific differences in oligodendrogenesis, and revealing areas of enhanced oligodendrocyte resilience and regenerative capacity following demyelination, demonstrating the utility of this pipeline for uncovering brain-wide oligodendrocyte dynamics in health and disease.

neuroscience↗

bioRxiv2024

Bi-level Graph Learning Unveils Prognosis-Relevant Tumor Microenvironment Patterns from Breast Multiplexed Digital Pathology

The tumor microenvironment is widely recognized for its central role in driving cancer progression and influencing prognostic outcomes. There have been increasing efforts dedicated to characterizing this complex and heterogeneous environment, including developing potential prognostic tools by leveraging modern deep learning methods. However, the identification of generalizable data-driven biomarkers has been limited, in part due to the inability to interpret the complex, black-box predictions made by these models. In this study, we introduce a data-driven yet interpretable approach for identifying patterns of cell organizations in the tumor microenvironment that are associated with patient prognoses. Our methodology relies on the construction of a bi-level graph model: (i) a cellular graph, which models the intricate tumor microenvironment, and (ii) a population graph that captures inter-patient similarities, given their respective cellular graphs, by means of a soft Weisfeiler-Lehman subtree kernel. This systematic integration of information across different scales enables us to identify patient subgroups exhibiting unique prognoses while unveiling tumor microenvironment patterns that characterize them. We demonstrate our approach in a cohort of breast cancer patients and show that the identified tumor microenvironment patterns result in a risk stratification system that provides new complementary information with respect to standard stratification systems. Our results, which are validated in two independent cohorts, allow for new insights into the prognostic implications of the breast tumor microenvironment. This methodology could be applied to other cancer types more generally, providing insights into the cellular patterns of organization associated with different outcomes.

bioinformatics↗

bioRxiv2023

Simultaneous multi-transient linear-combination modeling of MRS data improves uncertainty estimation

PurposeThe interest in applying and modeling dynamic MRS has recently grown. 2D modeling yields advantages for the precision of metabolite estimation in interrelated MRS data. However, it is unknown whether including all transients simultaneously in a 2D model without averaging (presuming a stable signal) performs similarly to 1D modeling of the averaged spectrum. Therefore, we systematically investigated the accuracy, precision, and uncertainty estimation of both described model approaches. MethodsMonte Carlo simulations of synthetic MRS data were used to compare the accuracy and uncertainty estimation of simultaneous 2D multi-transient LCM with 1D-LCM of the average. 2,500 datasets per condition with different noise representations of a 64-transient MRS experiment at 6 signal-to-noise levels for two separate spin systems (scyllo-inositol and GABA) were analyzed. Additional datasets with different levels of noise correlation were also analyzed. Modeling accuracy was assessed by determining the relative bias of the estimated amplitudes against the ground truth, and modeling precision was determined by standard deviations and Cramer-Rao Lower Bounds (CRLB). ResultsAmplitude estimates for 1D- and 2D-LCM agreed well and showed similar level of bias compared to the ground truth. Estimated CRLBs agreed well between both models and with ground truth CRLBs. For correlated noise the estimated CRLBs increased with the correlation strength for the 1D-LCM but remained stable for the 2D-LCM. ConclusionOur results indicate that the model performance of 2D multi-transient LCM is similar to averaged 1D-LCM. This validation on a simplified scenario serves as necessary basis for further applications of 2D modeling.

neuroscience↗

bioRxiv2022

Cross-modality supervised image restoration enables nanoscale tracking of synaptic plasticity in living mice

Synaptic plasticity encodes learning as changes in the strength of synapses, sub-micron structures that mediate communication between brain cells. Due to their small size and high density, synapses are extremely difficult to image in vivo, limiting our ability to directly relate synaptic plasticity with behavior. Here, we developed a combination of computational and biological methods to overcome these challenges. First, we trained a deep learning image restoration algorithm that combines the advantages of ex vivo super-resolution and in vivo imaging modalities to overcome limitations specific to each optical system. Applied to in vivo images from transgenic mice expressing fluorescently labeled synaptic proteins, this restoration algorithm super-resolved diffraction-limited synapses, enabling identification and logitudinal tracking of synaptic plasticity underlying behavior with unprecedented spatial resolution. More generally, our method demonstrates the capabilities of image enhancement to learn from ex vivo data and imaging techniques to improve in vivo imaging resolution.

neuroscience↗

bioRxiv2021

Antibody structure prediction using interpretable deep learning

Therapeutic antibodies make up a rapidly growing segment of the biologics market. However, rational design of antibodies is hindered by reliance on experimental methods for determining antibody structures. In recent years, deep learning methods have driven significant advances in general protein structure prediction. Here, we present DeepAb, a deep learning method for predicting accurate antibody FV structures from sequence. We evaluate DeepAb on two benchmark sets - one balanced for structural diversity and the other composed of clinical-stage therapeutic antibodies - and find that our method consistently outperforms the leading alternatives. Previous deep learning methods have operated as "black boxes" and offered few insights into their predictions. By introducing a directly interpretable attention mechanism, we show that our network attends to physically important residue pairs. For example, in prediction of one CDR H3 residue conformation, the network attends to proximal aromatics and a key hydrogen bonding interaction that constrain the loop conformation. Finally, we present a novel mutant scoring metric derived from network confidence and show that for a particular antibody, all eight of the top-ranked mutations improve binding affinity. These results suggest that this model will be useful for a broad range of antibody prediction and design tasks. SignificanceAccurate structure models are critical for understanding the properties of potential therapeutic antibodies. Conventional methods for protein structure determination require significant investments of time and resources and may fail. Although greatly improved, methods for general protein structure prediction still cannot consistently provide the accuracy necessary to understand or design antibodies. We present a deep learning method for antibody structure prediction and demonstrate improvement over alternatives on diverse, therapeutically relevant benchmarks. In addition to its improved accuracy, our method reveals interpretable outputs about specific amino acids and residue interactions that should facilitate design of novel therapeutic antibodies.

biophysics↗

bioRxiv2021

Automated in vivo tracking of cortical oligodendrocytes

Oligodendrocytes exert a profound influence on neural circuits by accelerating axon potential conduction, altering excitability and providing metabolic support. As oligodendrogenesis continues in the adult brain and is essential for myelin repair, uncovering the factors that control their dynamics is necessary to understand the consequences of adaptive myelination and develop new strategies to enhance remyelination in diseases such as multiple sclerosis. Unfortunately, few methods exist for analysis of oligodendrocyte dynamics, and even fewer are suitable for in vivo investigation. Here, we describe the development of a fully automated cell tracking pipeline using convolutional neural networks (Oligo-Track) that provides rapid volumetric segmentation and tracking of thousands of cells over weeks in vivo. This system reliably replicated human analysis, outperformed traditional analytic approaches, and extracted injury and repair dynamics at multiple cortical depths, establishing that oligodendrogenesis after cuprizone-mediated demyelination is suppressed in deeper cortical layers. Volumetric data provided by this analysis revealed that oligodendrocyte soma size progressively decreases after their generation, and declines further prior to death, providing a means to predict cell age and eventual cell death from individual time points. This new CNN-based analysis pipeline offers a rapid, robust method to quantitatively analyze oligodendrocyte dynamics in vivo, which will aid in understanding how changes in these myelinating cells influence circuit function and recovery from injury and disease.

neuroscience↗

bioRxiv2020

Geometric Potentials from Deep Learning Improve Prediction of CDR H3 Loop Structures

Antibody structure is largely conserved, except for a complementarity-determining region featuring six variable loops. Five of these loops adopt canonical folds which can typically be predicted with existing methods, while the remaining loop (CDR H3) remains a challenge due to its highly diverse set of observed conformations. In recent years, deep neural networks have proven to be effective at capturing the complex patterns of protein structure. This work proposes DeepH3, a deep residual neural network that learns to predict inter-residue distances and orientations from antibody heavy and light chain sequence. The output of DeepH3 is a set of probability distributions over distances and orientation angles between pairs of residues. These distributions are converted to geometric potentials and used to discriminate between decoy structures produced by RosettaAntibody. When evaluated on the Rosetta Antibody Benchmark dataset of 49 targets, DeepH3-predicted potentials identified better, same, and worse structures (measured by root-mean-squared distance [RMSD] from the experimental CDR H3 loop structure) than the standard Rosetta energy function for 30, 13, and 6 targets, respectively, and improved the average RMSD of predictions by 21.3% (0.48 [A]). Analysis of individual geometric potentials revealed that inter-residue orientations were more effective than inter-residue distances for discriminating near-native CDR H3 loop structures.

biophysics↗