The following sections provide details about the analyses performed in the paper entited:

Navigating the dynamic landscape of long noncoding RNA and protein-coding gene annotations in GENCODE

Saakshi Jalali, Shrey Gandhi, Vinod Scaria.

This document consists of detailed methodology alongwith all the associated content. We are only providing this information in addition to a description of methods as mentioned in the paper for making it easier to reproduce our analyses. Kindly note that this is not the release of a software package.

Data download

1. The annotated human transcriptome data was downlaoded from the GENCODE consortia through their ftp site (ftp://ftp.sanger.ac.uk/pub/gencode/Gencode_human)
2. The downloaded data was in form of gene transfer format (GTF) format.
3. The files for versions V1 to V24 were publically available for download. Note - V3a has not been made publically available.

Data Pre-processing

1. From the individual GTF files which were downloaded, we extracted all the transcripts, and made a list of unique transcript IDs assigned under the “transcript_id” identifier and their corresponding “havana_transcript” identifier if present. Uptil V3d the havana IDs were used as the main identifiers and thus we took them as such. Note - For V1, no gene or transcript information were available. Hence, we had to extract this information from the exonic coordinates present in the GTF file. We compiled the list of transcript biotype assigned to these unique transcripts which were extracted from GTF files (Compiled with all versions.txt)
2. For all ENSTs we removed transcript subcode numbers after the dot.
3. From the file, we replaced 218 ENST IDs with their respective ENSTR.
4. Removing redundancies by replacing OTTHUMT (Havana Ids) with ENST (Ensembl Ids):
   • 77,193 had single ENST prefixed (converted)
   • 1,982 OTTHUT prefixed IDs had more than one ENST IDs in the same version (OTTHUMT prefixed IDs were duplicated by assigning them both the Ensembl prefixed IDs while keeping their biotypes intact)
   • 3,188 OTTHUT prefixed IDs having more than one ENST prefixed IDs assigned to them across versions (ENST IDs duplicated and assigned biotypes of both the OTTHUMT IDs)
   • 3,272 as OTTHUMT as no ENST existed for them
5. Labeled transcript biotype as “NA” when no biotype was assigned in any particular version.

Processed Complete File

1. The complete Masterfile which we compiled consisted for 2,51,614 transcript ids.
2. We further assigned number codes to the biotypes present in the Masterfile (Masterfile converted to codes.txt). The assigned codes have been listed in the table below. ![Table] (https://github.com/vinodscaria/Gencode-moonlighting/blob/master/images/Picture4.png "Transcript biotypes")
3. We plotted the distribution of biotypes across all the GENCODE versions using RAW app as shown in the figure below. ![Pic] (https://github.com/vinodscaria/Gencode-moonlighting/blob/master/images/9.gif "Sankey")

Biotype wise data processing

1. Extracted the assigned biotypes for each transcript id across all GENCODE versions. Assigned “NA” to transcripts having no assigned biotype.
2. 1,14,114 protein coding transcripts were extracted across 28 versions (transcripts having "PC" biotype anytime in their lifetime)
3. Identified 23 sub-classes of lncRNAs as listed in Table below.
4. Renamed transcripts having anyone of the 23 biotype as lncRNA in our study.
5. 1,20,864 lncRNA transcripts extracted across 28 versions (transcripts having "any one of the 23" biotype anytime in their lifetime)
6. Compiled the total transcripts having PC and lncRNA biotype for further processing. ![Table2] (https://github.com/vinodscaria/Gencode-moonlighting/blob/master/images/Picture6.png "lncRNA biotypes")

To examine transcripts which are added or deleted in each version of GENCODE

1. To check number of added transcripts
   • If a biotype for a particular transcript was labelled as “NA” in Version n (Vn) and the same transcript had an assigned biotype (X) in next Version (Vn+1), then we considered that the transcript has been added in next version.
2. To check number of deleted transcripts
   • If a biotype for a particular transcript was assigned (X) in Version n (Vn) and the same transcript was labeled as “NA” in next Version (Vn+1), then we considered that the transcript has been deleted in next version. ![Table3] (https://github.com/vinodscaria/Gencode-moonlighting/blob/master/images/Picture1.png "Table Highlighting the addition and deletion of transcripts in each version")

Switching of transcripts across versions

1. If the biotype of a transcript in version n (Vn) is not equal to the biotype of same transcript in next version (Vn+1), then it was labelled as having moonlighting annotation. ![Table6] (https://github.com/vinodscaria/Gencode-moonlighting/blob/master/images/Picture5.png "Table Highlighting the transformation of transcripts across Gencode versions")

To check the consistency of biotype for a transcript

1. Transcripts having same biotype across all versions of GENCODE, we considered them to be consistent.
2. 19,520 lncRNA and 32,458 protein coding transcripts had same biotype across all GENCODE versions.
3. The transcripts which were inconsistent for their biotypes, were denoted to have moonlighting annotations.
4. We also calculated the number of biotypes which were assigned to any transcript over its lifetime. ![Table4] (https://github.com/vinodscaria/Gencode-moonlighting/blob/master/images/Picture2.png "Table Highlighting the number of biotypes assigned to transcript during its lifetime")
5. In addition, we also observed transcripts went through large number of transitions from one biotype to another in its lifetime. The same has been listed below. ![Table5] (https://github.com/vinodscaria/Gencode-moonlighting/blob/master/images/Picture3.png "Table Highlighting the number of transitions each transcript went through during its lifetime")