"3DisoTransFold" pipeline was used to evaluate translation ORFs and predict isoform structures across transcriptome-identified transcripts, ultimately constructing a protein structural similarity network, “3DisoGalaxy,” based on multiple structure alignment scores comparing isoform structures. Specifically, the predicted ORFs from the master transcriptome were retained only if identified by Ribo-seq analysis, supporting their participation in the translation process.
By combining BLASTP and HMMER methods, ORF sequences were predicted using TransDecoder and machine learning. After filtering out low-confidence and short ORFs, the longest ORF sequence was kept for each isoform, and AlphaFold-2.3 was utilized to predict protein structures. Structural similarity was then calculated using the TMalign method, collecting pairwise isoform similarities to construct a “protein universe.” Additionally, functional domains and GO term pathway annotations were assigned to the network based on UniProtKB databases.
The pipeline 
https://github.com/CityUHK-CompBio/TranStructomics, was used to evaluate translation ORFs and predict isoform structures across transcriptome-identified transcripts, ultimately constructing a protein structural similarity network, “3DisoGalaxy,” based on multiple structure alignment scores comparing isoform structures. Specifically, the predicted ORFs from the master transcriptome were retained only if identified by Ribo-seq analysis, supporting their participation in the translation process.
By combining BLASTP and HMMER methods, ORF sequences were predicted using TransDecoder and machine learning. After filtering out low-confidence and short ORFs, the longest ORF sequence was kept for each isoform, and AlphaFold-2.3 was utilized to predict protein structures. Structural similarity was then calculated using the TMalign method, collecting pairwise isoform similarities to construct a “protein universe.” Additionally, functional domains and GO term pathway annotations were assigned to the network based on UniProtKB databases.
Through the final network, the landscape of cancer-specific alternative splicing-induced protein isoforms becomes available, allowing for the screening and validation of subtype-specific isoforms for therapeutic applications.
A case study version of the final network, “3DisoGalaxy”, can be accessed at
and the “3DisoTransFold” pipeline for analyzing all data is available at