Garg, D. | bioRxiv Science

bioRxiv2026

Deep Brain Stimulation Microelectrodes as a Source of Human Subcortical RNA: Validation of a Low-Input Transcriptomic Protocol

BackgroundUnderstanding the molecular basis of Parkinsons disease (PD) phenotypic heterogeneity maybe improved by in vivo access to deep brain tissue. Deep brain stimulation (DBS) surgery offers a unique opportunity: microelectrodes traversing the subthalamic nucleus (STN) carry adherent brain tissue upon withdrawal, providing a source of RNA from subcortical regions in living patients without additional invasive procedures. AimTo develop and validate a low-input RNA extraction and transcriptomic profiling protocol using DBS microelectrodes in post-mortem human brain, to confirm that recovered RNA is of brain rather than blood origin, and to characterise its sub-regional and cellular identity. MethodologyDBS microelectrodes were inserted without image guidance or guide tube into three unfixed post-mortem human brains targeting the STN trajectory. In clinical practice, a guide tube shields the electrode from cortical tissue; its absence here means tissue from the full insertion trajectory may contribute to recovered RNA. Thirty-eight microelectrodes were evaluated across single and pooled strategies. RNA was extracted using a modified RNeasy Micro low-input protocol; libraries prepared using NEBNext Single Cell/Low Input RNA Library Kit and sequenced on Illumina NovaSeq 6000 (paired-end, 2x150 bp). Tissue identity was validated against GTEx v10 (54 tissues) and Allen Human Brain Atlas (ABA, 19 subcortical regions including STN) using Pearson correlation with permutation testing (1,000 permutations) and BH-FDR correction. Transcriptional overlap between subcortical and cortical reference regions was quantified and subcortical-enriched gene filtering performed. ResultsTwenty-five RNA isolates were obtained from 38 microelectrodes; 54.5% of Bioanalyzer-assessed samples achieved RIN [≥]5 (median 7.1; range: 5.9-7.8). Fifteen libraries passed sequencing QC (mean depth 43.0 {+/-} 14.1 million read pairs; mean Q30 83.2 {+/-} 5.6 %). PCA of rlog transformed expression data resolved samples by donor identity. All samples confirmed brain tissue origin by GTEx {tau}-index tissue specificity correlation (n=996 brain and blood specific markers). Brain correlations were highest for frontal cortex (mean r= 0.683 {+/-} 0.072), anterior cingulate cortex (mean r= 0.658 {+/-} 0.069), amygdala (mean r= 0.615 {+/-} 0.068), and basal ganglia including caudate (mean r= 0.553 {+/-} 0.059), putamen (mean r= 0.546 {+/-} 0.059), and nucleus accumbens (mean r= 0.544 {+/-} 0.057); all BH-FDR < 0.05. Cerebellum showed the lowest brain correlations (cerebellar hemisphere: mean r = 0.327 {+/-} 0.039; cerebellum: mean r = 0.314 {+/-} 0.037). Whole blood correlation was strongly negative (mean r= -0.250 {+/-} 0.079), confirming non-haematological origin. ABA genome-wide analysis confirmed positive STN correlation (r=0.54-0.62; permutation p<0.001). MuSiC cell-type deconvolution against the Allen Brain Atlas HMBA-BG snRNA-seq reference identified oligodendrocytes (33.5 {+/-} 17.2%), frontal cortical neurons (9.1 {+/-} 7.0%), STR D1 MSNs (6.0 {+/-} 4.9%), and dopaminergic neurons (1.1{+/-} 2.7%) as the principal cell types, independently corroborating bulk transcriptomic findings at single-cell resolution. ConclusionThis study validates a low-input RNA extraction protocol for DBS microelectrodes, confirming brain-specific transcriptomic profiles consistent with STN-adjacent subcortical sampling. The 96% transcriptional overlap between cortical and subcortical regions, combined with absence of a guide tube in this post-mortem model, limits sub-regional specificity; clinical application with a guide tube would enrich the subcortical signal. These findings provide the methodological framework for in vivo molecular profiling of the human basal ganglia during DBS surgery in PD patients.

genetics↗

bioRxiv2025

CPI-Pred: A deep learning framework for predicting functional parameters of compound-protein interactions

Recent advancements in deep learning have enabled functional annotation of genome sequences, facilitating the discovery of new enzymes and metabolites. However, accurately predicting compound-protein interactions (CPI) from sequences remains challenging due to the complexity of these interactions and the sparsity and heterogeneity of available data, which constrain the generalization of patterns across their solution space. In this work, we introduce CPI-Pred, a versatile deep learning model designed to predict compound-protein interaction function. CPI-Pred integrates compound representations derived from a novel message-passing neural network and enzyme representations generated by state-of-the-art protein language models, leveraging innovative sequence pooling and cross-attention mechanisms. To train and evaluate CPI-Pred, we compiled the largest dataset of enzyme kinetic parameters to date, encompassing four key metrics: the Michaelis-Menten constant (KM), enzyme turnover number (kcat), catalytic efficiency (kcat/KM), and inhibition constant (KI).These kinetic parameters are critical for elucidating enzyme function in metabolic contexts and understanding their regulation by compounds within biological networks. We demonstrate that CPI-Pred can predict diverse types of CPI using only the amino acid sequence of enzymes and structural representations of compounds, outperforming state-of-the-art models on unseen compounds and structurally dissimilar enzymes. Over workflow provides a valuable tool for tackling a range of metabolic engineering challenges, including the designing of novel enzyme sequences and compounds, such as enzyme inhibitors. Additionally, the datasets curated in this study offer a valuable resource for the scientific community, serving as a benchmark for machine learning models focused on enzyme activity and promiscuity prediction.

bioinformatics↗