Biology subjects

Guruacharya, A.

Publications and source records attributed to Guruacharya, A..

7 recordsLinked to original sources

bioRxiv2026

AI-guided design and ex vivo validation of nanobodies targeting aggregation motifs of intrinsically disordered protein tau

Intrinsically disordered proteins (IDPs) represent major yet challenging therapeutic targets in neurodegenerative disease due to their conformational heterogeneity and aggregation-prone behavior. Tau protein is a prototypical IDP that forms pathological aggregates in Alzheimers disease and related tauopathies. Despite extensive clinical efforts, tau-directed monoclonal antibodies have demonstrated limited efficacy. Concurrently, single-domain antibodies (nanobodies) have been gaining importance because of their small size and membrane penetrating capabilities. New design paradigms are therefore required for nanobodies to enable precise targeting of disease-relevant conformations. Here, we describe a biophysical modelling and AI-guided nanobody discovery targeting the VQIVYK motif of tau, which constitutes the structural core of neurofibrillary tangles in Alzheimers Disease. Biophysical modelling-based target analysis identified low-energy conformational states of VQIVYK. These conformational insights were used to guide AI-driven nanobody design of CDR3 loops. Starting from a nanobody scaffold, we generated 145 candidate nanobodies through systematic backbone sampling and neural network-guided sequence design, followed by multi-dimensional computational prioritization. Two candidates demonstrated robust binding to synthetic full tau protein in ELISA binding assays, achieving binding indices of 148.9% and 140%, relative to reference controls. Notably, one candidate also exhibited strong reactivity in post-mortem Alzheimers disease human brain tissue, with a binding index of 236.1%, exceeding that of the positive control (222.9%). Structural analysis indicates that our nanobodies engineered CDR3 engages VQIVYK through optimized aromatic and hydrophobic interactions. Together, these findings establish a proof-of-concept for biophysics-guided, AI-guided nanobody engineering against IDPs and identifies them as a promising lead for tau-targeted single domain antibody development.

neuroscience↗

bioRxiv2026

Cirrina: LLM-driven pharmacological reasoning agent enables preclinical CNS drug evaluation

Assessing whether a preclinical drug candidate will work is not a prediction problem but a reasoning problem. The same numerical output warrants different interpretations depending on the target and therapeutic context. CNS drug development presents the most demanding instance of this reasoning problem. For example, a compound must cross the blood-brain barrier, resist efflux transport, and achieve adequate receptor occupancy at a dose that clears safety margins. The constraints interact with each other in a web that needs careful interpretation. Here, we show that Cirrina, an LLM agent coupled to eight mechanistic pharmacology tools, can reason across the input data to provide better decisions and a well documented reasoning trace. The LLM agent reasons across multiple data tiers from SMILES to animal PK/PD measurements adjusting thresholds based on target-specific requirements. Validated against 181 CNS compounds, it achieved a 68% accuracy compared to a rule-based deterministic pipeline of 31% accuracy. In 103 discordant cases, the agents reasoning was correct in 75% of instances compared to only 10% for deterministic pipelines. Cirrina provides a scalable, documented framework for preclinical decision-making, effectively identifying failure-prone candidates that generic thresholds overlook, and thereby reducing the chances of failure along the clinical development cycle.

pharmacology and toxicology↗

bioRxiv2026

Translational Bayesian Pharmacokinetic Framework for Uncertainty-Aware First-in-Human Dose Selection of Therapeutic Monoclonal Antibodies

First-in-human (FIH) dose selection for monoclonal antibodies (mAbs) typically relies on deterministic allometric scaling but lacks formal uncertainty quantification. While Bayesian methods have been widely applied in population PK modeling and dose individualization, their use for propagating uncertainty through allometric scaling in mAb FIH dose selection has not been systematically explored. This is a critical limitation for molecules with narrow therapeutic windows, such as CNS-targeted mAbs, where the blood-brain barrier restricts IgG penetration to [~]0.1-0.3% of plasma concentrations, requiring high systemic doses that must be balanced against dose-limiting toxicities. To provide uncertainty-aware FIH dose recommendations, we developed and systematically evaluated a Bayesian hierarchical PK framework tested on CNS mAbs. By simultaneously learning population-level PK distributions and allometric scaling relationships from 9 well-characterized mAbs, the model propagates inter-antibody variability and scaling imprecision into full posterior predictive distributions. For validation, the framework was applied to 3 Alzheimers disease mAbs, donanemab, lecanemab, and aducanumab, using only cynomolgus monkey PK data to predict human outcomes. Leave-one-out cross-validation yielded a mean absolute prediction error of 11.6% for human clearance. Predicted FIH doses were 10 mg/kg for donanemab and lecanemab, and 30 mg/kg for aducanumab. Retrospective comparison with clinical data showed prediction errors of -36.1%, -36.1%, and -15.7%, respectively, all within two-fold of observed values. The systematic under-prediction of clearance is attributable to target-mediated drug disposition not captured by the linear model. However, this bias is pharmacologically conservative, as it over-predicts systemic exposure to ensure wider safety margins. This framework enables risk-informed FIH dose selection by providing full probability distributions of predicted exposures rather than point estimates.

pharmacology and toxicology↗

bioRxiv2025

Quantum Computing Reveals Energetics of Tau Peptide Fragments

Near-term quantum algorithms such as the variational quantum eigensolver (VQE) have been widely explored for small-molecule electronic structure calculations, yet their relevance for biologically motivated peptide systems remains largely untested. Here, we apply a rigorously controlled, fragment-based VQE workflow to a tau-derived peptide fragment implicated in protein aggregation in Alzheimers Disease. Using an identical active space, basis set, and frozen-core treatment, we benchmark VQE electronic energies against classical restricted Hartree-Fock (RHF) calculations and molecular dynamics (MD) force-field energies across an ensemble of peptide conformations. While VQE and active-space RHF energies show systematic agreement within the defined electronic subspace, both exhibit weak correlation with MD-derived energetics, highlighting the fundamentally different physical contributions captured by electronic structure methods and classical force fields. These results demonstrate that NISQ-era quantum chemistry provides complementary, rather than redundant, information relative to classical MD and delineate the scope and limitations of applying VQE to biologically relevant peptide fragments. Our study establishes a disease-motivated benchmark framework for integrating quantum electronic structure calculations with classical simulation approaches in peptide biophysics.

biophysics↗

bioRxiv2025

codonGPT: Reinforcement learning on a generative language model optimizes RNA sequences under biological constraints

Emerging generative models for biology focus on DNA, non-coding RNA, or proteins, ignoring information hidden in mRNA. Additionally, in protein engineering and mRNA therapeutics the design of mRNA sequences is still a challenge, lacking a clear framework. Here, we introduce and rigorously evaluate two novel methods: a foundational model for mRNA and a reinforcement learning mRNA design framework built on such a model. codonGPT is the first generative foundational language model trained directly on coding mRNA sequences. To solve the problem of synonymous constraints that are only unique to mRNA, we introduce a novel method of inference-time masking, along with house-keeping genes evaluation. For the first time, we also rigorously demonstrate, that for precise mRNA therapeutics design, reinforcement learning on such a model provides a clear framework for biological optimization. Our study introduces a novel foundational model for mRNA and a new reinforcement learning based paradigm for mRNA sequence design.

bioinformatics↗

bioRxiv2023

Co-evolution of ion channels and neurotoxins in cnidarians leads to diversification of ion channel genes

Understanding the diversity of ion channels in cnidarians may shed light on the origin and evolution of early nervous systems. It is hypothesized that cnidarian neurotoxins led to the evolution of diverse ion channel proteins in cnidarians. I tested this hypothesis by investigating several evolutionary factors of both cnidarian neurotoxins and their target ion channels. I examined homologs of 250 cnidarian toxins, 75 ion channel genes, and 70 housekeeping genes from 37 transcriptomes/genomes of cnidarian species. Analysis on the presence or absence of genes by species, selection analysis, and evolutionary rate analysis were performed on the homologs of neurotoxin and ion channel proteins. I found evidence of positive selection, correlation between the number of homologous gene families, and difference in the evolutionary rates among the gene families. I have shown for the first time that neurotoxins may have coevolved with the ion channels in cnidarians. This is consistent with an evolutionary arms race between ion channels and neurotoxins leading to extensive diversity of ion channel genes found in cnidarians.

evolutionary biology↗

bioRxiv2018

Panama: a tool for ion channel biophysics simulation

I have created an online tool and an R library that simulates biophysics of voltage-gated ion channels. It is made publicly available as an R library called Panama at github.com/anuj2054/panama and as a web app at neuronsimulator.com. A need for such a tool was observed after surveying available software packages. I found that the available packages are either not robust enough to simulate multiple ion channels, too complicated, usable only as desktop software, not optimized for mobile devices, not interactive, lacking intuitive graphical controls, or not appropriate for undergraduate education. My app simulates the physiology of 11 different channels - voltage-gated sodium, potassium, and chloride channels; channels causing A-current, M-current, and After-HyperPolarization (AHP) current; calcium-activated potassium channels; low threshold T type calcium channels and high threshold L type calcium channels; leak sodium and leak potassium channels. It can simulate these channels under both current clamp and voltage clamp conditions. As we change the input values on the app, the output can be instantaneously visualized on the web browser and downloaded as a data table to be further analyzed in a spreadsheet program. The app is a first of its kind, mobile-friendly and touch-screen-friendly online tool that can be used to teach undergraduate neuroscience classes. It can also be used by researchers on their local computers as part of an R library. It has intuitive touch-optimized controls, instantaneous graphical output, and yet is pedagogically robust for education and casual research purposes.\n\nNeuroscience education, ion channel biophysics, Hodgkin-Huxley simulation, web app for neuroscience

bioinformatics↗