Biology subjects

Banerjee, C.

Publications and source records attributed to Banerjee, C..

6 recordsLinked to original sources

bioRxiv2026

Conformalized Multiview Learning

SO_SCPLOWUMMARYC_SCPLOWThere is an overwhelmingly large body of literature and numerous algorithms already available on "multimodal AI", based on different modeling and fusion paradigms. As multimodal AI becomes fully integrated into healthcare applications, the ability to say "Im not sure" or "I dont know" when uncertain is a necessary capability for safe clinical deployment, requiring the augmentation of current multimodal AI techniques. Our goal in this work is not to add yet another method to the growing toolbox of multimodal AI, but rather to (i) clarify how existing supervised multimodal AI methods can be understood through the lens of uncertainty quantification, and (ii) develop a unified framework that provides valid and interpretable uncertainty estimates for multimodal AI practitioners. We introduce Coracle, a conformalized framework for multimodal AI that seamlessly adapts to early, late, and intermediate fusion. Coracle provides theoretical marginal confidence guarantees and achieves valid finite-sample coverage without relying on distributional assumptions. Through extensive simulations and analyses of diverse biomedical multiview datasets, we demonstrate that Coracle consistently outperforms existing non-conformalized methods in both uncertainty estimation and calibration. By equipping published multiview models with valid, interpretable confidence measures, Coracle advances the development of trustworthy, uncertainty-aware statistical tools for biomedical decision support. Coracle is publicly available as an open-source R package at https://github.com/himelmallick/Coracle.

systems biology↗

bioRxiv2025

Repurposing without Reinventing the Wheel - Ensemble Models for Differential Analysis

Inspired by ensemble models in machine learning, we propose a general framework for aggregating multiple distinct base models to enhance the power of published differential association analysis (DAA) methods. We demonstrate this approach by augmenting popular DAA models with one or more biologically motivated alternatives. This creates an ensemble that bypasses the challenge of selecting an optimal model and instead combines the strengths of complementary statistical models to achieve superior performance. Our proposed ensemble learning approach is platform-agnostic and can augment any existing DAA method, providing a general and flexible framework for various downstream modeling tasks across domains and data types. We performed extensive benchmarking across both simulated and experimental datasets spanning single-cell gene expression, bulk transcriptomics, and microbiome metagenomics, where the ensemble strategy vastly outperformed non-ensemble methods, identified more differential patterns than the competing methods, and displayed good control of false positive and false discovery rates across diversified scenarios. In addition to highlighting a substantial performance boost for state-of-the-art DAA methods, this work has practical implications for mitigating the so-called reproducibility crisis in omics data science. An open-source R package implementing the ensemble strategy is publicly available at https://github.com/himelmallick/DAssemble.

bioinformatics↗

bioRxiv2023

Efficient Prohibitin 2 exposure during mitophagy depends on Voltage-dependent anion-selective channel protein 1.

Autophagic elimination of depolarized mitochondria (mitophagy) depends on Ubiquitin proteasome complex to expose the inner mitochondrial membrane-resident protein-Prohibitin 2 (PHB2). This uncovering facilitates its interaction with autophagosomal membrane-associated protein LC3. It remains unclear whether PHB2 is uncovered randomly through mitochondrial rupture sites. Prior knowledge and initial screening indicated that Voltage-dependent anion-selective channel protein 1 (VDAC1) might play a role in this process. Through in vitro biochemical assays and imaging, we have found that VDAC1-PHB2 interaction increases during mitochondrial depolarization. Subsequently, this interaction enhances the efficiency of PHB2 exposure and mitophagy. To investigate the relevance in vivo, we utilized a Porin (equivalent to VDAC1) knockout Drosophila line. Our findings demonstrate that during rotenone-induced mitochondrial stress, Porin is essential for PHB2 exposure, PHB2-LC3 interaction, and mitophagy. This study highlights that VDAC1 predominantly synchronizes efficient PHB2 exposure through mitochondrial rupture sites during mitophagy. These findings may provide insights to understand progressive neurodegeneration.

cell biology↗

bioRxiv2023

Structural basis of TFIIIC-dependent RNA Polymerase III transcription initiation

RNA Polymerase III (Pol III) is responsible for transcribing 5S ribosomal RNA (5S rRNA), tRNAs, and other short non-coding RNAs. Its recruitment to the 5S rRNA promoter requires transcription factors TFIIIA, TFIIIC, and TFIIIB. Here we use cryo-electron microscopy to visualize the S. cerevisiae complex of TFIIIA and TFIIIC bound to the promoter. Brf1-TBP binding further stabilizes the DNA, resulting in the full-length 5S rRNA gene wrapping around the complex. Our smFRET study reveals that the DNA undergoes both sharp bending and partial dissociation on a slow timescale, consistent with the model predicted from our cryo-EM results. Our findings provide new insights into the mechanism of how the transcription initiation complex assembles on the 5S rRNA promoter, a crucial step in Pol III transcription regulation.

biophysics↗

bioRxiv2021

Characterizing Locus Specific Chromatin Structure and Dynamics with Correlative Conventional and Super Resolution imaging in living cells

The dynamic rearrangement of chromatin is critical for gene regulation, but mapping both the spatial organization of chromatin and its dynamics remains a challenge. Many structural conformations are too small to be resolved via conventional fluorescence microscopy and the long acquisition time of super-resolution PALM imaging precludes the structural characterization of chromatin below the optical diffraction limit in living cells due to chromatin motion. Here we develop a correlative conventional fluorescence and PALM imaging approach to quantitatively map time-averaged chromatin structure and dynamics below the optical diffraction limit in living cells. By assigning localizations to a locus as it moves, we reliably discriminate between bound and searching dCas9 molecules, whose mobility overlap. Our approach accounts for changes in DNA mobility and relates local chromatin motion to larger scale domain movement. In our experimental system, we show that compacted telomeres have a higher density of bound dCas9 molecules, but the relative motion of those molecules is more restricted than in less compacted telomeres. Correlative conventional and PALM imaging therefore improves the ability to analyze the mobility and time-averaged nanoscopic structural features of locus specific chromatin with single molecule precision and yields unprecedented insights across length and time scales.

biophysics↗

bioRxiv2020

Quantitative super-resolution microscopy reveals distinct ULK1 oligomeric states and nanoscopic morphologies during autophagy initiation

Autophagy induction involves extensive molecular and membrane reorganization. Despite significant progress, the mechanism underlying autophagy initiation remains poorly understood. Here, we used quantitative Photoactivated Localization Microscopy with single-molecule sensitivity to analyze the nanoscopic distribution of endogenous ULK1, the kinase that triggers autophagy. Under amino acid starvation, ULK1 formed large clusters containing up to 161 molecules at the endoplasmic reticulum. Cross-correlation analysis revealed that ULK1 clusters engaging in autophagosome formation require 30 or more molecules. The ULK1 structures with more than the threshold number contained varying levels of Atg13, Atg14, Atg16, LC3B, GEC1 and WIPI2. We found that ULK1 activity is dispensable for the initial clustering of ULK1, but necessary for the subsequent expansion of the clusters, which involves interaction with Atg14, Atg16, and LC3B and relies on Vps34 activity. This quantitative analysis at the single molecule level has provided unprecedented insights into the behavior of ULK1 during autophagy initiation.

biophysics↗