We used #eQTL analysis to examine how SNPs relate to gene expression in #Habenula, 7 eQTLS genes were identified in differential expression, and 16 eSNPs were schizophrenia GWAS risk SNPs. Nine new schizophrenia colocalized genes with genetic risk were identified

We investigated whether the DEGs we found were unique to the #Habenula vs. other brain regions studied in schizophrenia (Caudate, Dentate Gyrus, dlPFC, Hippocampus) and found 75% of our DEGs were unique to habenula 🦄

Differential gene expression yielded 173 differentially expressed genes between schizophrenia and control donors, some of which support the neurodevelopmental hypothesis of schizophrenia 🌋

To refine habenula signal from neighboring thalamus we utilized our #snRNAseq data to perform #deconvolution on our bulk samples, estimating proportion of habenula. These proportions were used to control for dissection differences in the differential expression analysis

We also profiled habenula-enriched tissue from schizophrenia and control donors using bulk RNA-seq #Transcriptomics, generating a unique dataset to identify disease associated molecular changes. PCA with bulk data from other regions shows #Habenula as an outlier

To provide cross-species context we integrated our human Hb snRNA-seq dataset with a previously published mouse #Habenula #singlecell dataset from @gstuber.bsky.social. Some neuronal subclusters such as human OPRM1-enriched LHb.2 appeared to be conserved across species 🐁

Single molecule fluorescence in situ hybridization #RNAScope and HALO from @indicalabs.bsky.social was used to visualize the spatial organization of these human Hb neuronal populations. 🔬 For example, MHb.2 neurons that highly expressed CHAT clustered separately from other MHb cell types

We’ve created a #snRNAseq molecular atlas of the human Habenula to understand neuronal diversity on a molecular level in this unique brain region. Our snRNA-seq data on 7 control donors yielded 16k nuclei and 17 clusters, including 7 lateral & 3 medial Hb neuron populations 🧪

Thank you to all of our collaborators at the @lieberinstitute.bsky.social for their dedication and hard work on this project! 🧑‍🔬 #HabenulaLIBD @freneegf.bsky.social @nick-eagles.bsky.social @lcolladotor.bsky.social @kr-maynard.bsky.social @martinowk.bsky.social

We also propose and evaluate a new method for selecting cell type specific #markergenes for deconvolution: #MeanRatio implemented in #DeconvoBuddies
@bioconductor.bsky.social #rstats
🛠️ research.libd.org/DeconvoBuddi...

To assess method accuracy, we compared the estimated proportions to the #RNAScope proportions for correlations and rmse 📈 #BisqueRNA and #hspe were the top performers 🏆

We ran #deconvolution on the 110 bulk RNA-seq samples, with six selected methods: #DWLS, #bisqueRNA, #MuSiC, #BayesPrism, #hspe, and #CIBERSORTx. The methods returned a wide range of proportion estimates 📊

The resulting cell type proportions had a large range across the different #DLPFC samples, and were not always correlated with the paired #snRNAseq proportions ⚠️

We designed an #RNAScope/IF experiment to measure six broad cell types in the #DLPFC. We used #HALO from Indica Labs to count the occurrence of the cell types across the tissue sections 🔬 #ImmunoFlourescence (awesome work from Kelsey Montgomery)

Testing these #deconvolution methods was made possible by a really cool multi-assay dataset from consecutive slices of #DLPFC tissue blocks 🧠1. Bulk RNA-seq 🧬 2. #snRNAseq as the reference 🦠 3. #RNAScope serving as a orthogonal measurement of cell type proportions 🔬