DatasetsWe analyzed publicly available in vivo paired-end RNA sequencing data of the mouse cerebral cortex and hippocampus as well as in vivo data from neural progenitor cells (NPCs) and neurons (Fig. 1c) with a minimum read length of 100 bp. We downloaded the sequencing data from NCBI SRA or GEO respectively. Specifically, we used SRP055008 [6], GSE133291 [22], and GSE96950 [9]. Further datasets can easily be integrated, refer to https://cortexa-rna.com/datasets.RNA-seq analysisWe used a standardized RNA-seq pipeline (Fig. 1a) to analyze the transcriptomic data for gene expression and alternative splicing. In brief, we trimmed the reads for adapter sequences with BBDuk (version 39.01) [30]. The trimmed reads were aligned to the reference genome mm39 (released 19.10.2022) downloaded from Gencode using STAR (version 2.7.10b) [31] and indexed using samtools (version 1.18) [32]. FeatureCounts, provided by SubRead (version 2.0.6) [33], was used to quantify the expression of each respective gene. All gene expression counts were normalized to transcripts per million (TPM).We utilized rMATS turbo (version 4.1.2) [34] with default settings to detect AS events. We analyzed the data for five alternative splicing events: cassette exon (skipping exon), mutually exclusive exons, intron retention, alternative A5′ splice site, and alternative A3′ splice site (Fig. 1b), as defined in rMATS [34]. The coverage for sashimi plots was analyzed in 10 bp steps with bamCoverage (version 3.5.2) from deepTools2 [35] converted to wig with Encode bigWigToWig.WebappThe web application was built with Next.js frontend framework, utilizing SQLite database for backend data storage. TypeScript enhances code maintainability and type safety, while Tailwind CSS streamlined styling. Prisma serves as the ORM tool for efficient database management. The application follows the REST principles for communication between frontend and backend components, optimizing interoperability and scalability. The website is hosted on servers of the Johannes Gutenberg University, Mainz, Germany. A detailed tutorial on how to interpret alternative splicing events presented in Cortexa is provided at https://cortexa-rna.com/tutorial.Visualization of genesGene expression was normalized to transcripts per million (TPM) and represented as a barplot for each dataset. Alternative splicing events (Fig. 1b) are visualized as sashimi plots.SplicePCASplicePCA performs principal component analysis (PCA) on averaged percentage spliced-in (PSI) values (Fig. 2a). Initially, the user can either select a subset of genes or perform SplicePCA on all genes. All events that have missing values are removed from the subsequent analysis. Next, PSI values are averaged within their respective group (e.g. E14.5 from the developmental data set [6]). These averaged PSI values are the input for the PCA, and the resulting values are plotted and available for download. Moreover, SplicePCA allows users to integrate their in-house analyzed output files (rMATS). We recommend processing the files as described in Sect. 3.2 and in the tutorial available at https://github.com/s-weissbach/cortexa_SplicePCA_example/.Fig. 2a Schematic representation of SplicePCA. SplicePCA takes percentage spliced-in (PSI) values for all genes or a selected subset of genes as input. Alternative spicing events with missing values are removed from the dataset. The remaining PSI values are averaged over individual replicates per experimental group. Finally, PCA is performed on the averaged data, resulting in a representation of splicing patterns across groups in two-dimensional space. b Cassette exon data of cortical samples of WT (E18.5) and Nova2-KO (E18.5), analyzed with SplicePCA and compared to splicing changes across normal development of the mouse cortex. PCA of alternative splicing data across cortical development forms a characteristic bell-shaped trajectory (indicated by gray arrows) with P0 as its inflection point. The conditional knock-out of Nova2 resulted in a substantial deviation from the inferred normal splicing trajectory. Nova2-KO samples (E18.5) were associated with a less mature splicing pattern than E16.5 wild-type samplesExample usage of SplicePCATo demonstrate the use of SplicePCA, we obtained Nova2- knock-out (KO) and wild-type (WT) data from NCBI GEO with the accession number GSE103314 [36]. We performed quality control, trimming, alignment, and alternative splicing analysis as described in Sect. 3.2 RNA-seq analysis.Subsequently, the cassette exons from the rMATS output file were uploaded to https://cortexa-rna.com/pca and analyzed in the context of developmental [6] and NPC/neuron-specific [9] alternative splicing events. Next, the results from SplicePCA were downloaded and plotted using matplotlib (version 3.9.0) [37].Utility and discussionAlternative splicing is a prevalent regulatory mechanism in the brain that plays an important role during development and in specifying and maintaining neural cell types [6, 7, 9, 22, 28, 36, 38]. However, Mus musculus has ~ 22,000 protein-coding genes [39] of which almost all multi-exon genes undergo alternative splicing [40]. Functional implications of these alternative splicing events remain in many cases elusive. Cortexa is an easy-to-use web tool to access alternative splicing events for genes of interest in a developmental and neuronal cell-type-specific context to formulate and investigate research hypotheses.Additionally, principal component analysis has proven to be a powerful tool to summarize alterations in the alternative splicing landscape [6, 27,28,29]. Representation of samples in two-dimensional space allows investigating the similarity or divergence of global splicing patterns between different experimental conditions. In our analysis of developmental alternative splicing, we observed a characteristic bell-shaped trajectory across diverse iterations, which aligns with findings reported in the literature [6]. However, the interpretation of principal components can be challenging in terms of associating them with biologically meaningful factors. By design, principal components capture the direction of maximal variance tin the original data [41] which does not necessarily reflect experimental or biological factors. Despite these limitations, PCA remains a valuable exploratory tool, and SplicePCA offers a user-friendly method for investigating alternative splicing in the context of development or different cell types.Using SplicePCA, researchers can select splice events for specific genes, integrate their data, and interpret results in the context of cortical development and specific neuronal cell types. To showcase the usefulness of this approach, we re-analyzed cortical Nova2-KO and WT samples at embryonic day E18.5 [36] and used the SplicePCA tool. NOVA2 belongs to the class of RNA-binding proteins, governing alternative splicing during cortical development and in mature neurons [42]. Specifically, NOVA2 is required to regulate neuronal migration through splicing Dab1 which is part of the Reelin pathway [2]. By using SplicePCA, we revealed a striking effect of Nova2-KO during cortical development. E18.5 knockout samples were located between E14.5 and E16.5 wild-type samples on the inferred developmental splicing trajectory, indicating a less mature splicing pattern (Fig. 2b). Thus, NOVA2 splicing activity contributes significantly to the splicing changes between E14.5 to E18.5, as reported previously [2, 36, 43]. These results support the relevance of SplicePCA which combines available datasets with new datasets. Cortexa thus allows the use of publicly available data without extensive re-analysis, which would otherwise require significant computational resources.
