Other scripts used in this study

Generation of the GTEx gene model

We used TransDecoder (version 5.5.0) was used to find the coding sequences from the GTEx model.

The GTEx model can be found in the Zenodo.

To eliminate potential transcriptional noise, transcripts with the max Transcripts Per Million (TPM) values ≥ 2 among all GTEx RNA-seq samples and detected in ≥ 3 samples were kept.

Transcript sequences were acquired with the “gtf_genome_to_cdna_fasta.pl” tool in TransDecoder.

util/gtf_genome_to_cdna_fasta.pl $FILE_GTF $FILE_GENOME >$FILE_TRANSCRIPT

The sample scripts for protein translation were shown below.

TransDecoder.LongOrfs -m 30 -t $FILE_TRANSCRIPT
TransDecoder.Predict -t $FILE_TRANSCRIPT

The genomic location of proteins were determined with the script below.

util/cdna_alignment_orf_to_genome_orf.pl $FILE_TRANSCRIPT.transdecoder.gff3 $FILE_GTF $FILE_TRANSCRIPT > $FILE_TRANSCRIPT.transdecoder.genome.gff3

where

• FILE_GTF: the GTF file
• FILE_GENOME: the genome file
• FILE_TRANSCRIPT: the transcript file. It is also used as the output prefix for TransDecoder results, such as $FILE_TRANSCRIPT.transdecoder.gff3.

Proteins aligned to the genome by TransDecoder were combined for further analysis. We use CD-HIT to identify proteins which were identical or part of longer proteins. The command is shown below.

cd-hit -i $FILE_PROTEINS_COMBINED -o $OUTPUT_FILE_DEDUP_PROTEINS -c 1.0 -n 5 -M 250000 -T 28 -U 0 -s 0 -uL 0 -uS 0

Translation of GTEx model directly

The GTEx model was also translated directly with TransDecoder. The minimum protein length was set to 30.

blast search

Sample scripts of how the homologous search was performed.

nr

• The NCBI nr sequences were split to 500 parts.
• DIAMOND (v2.0.11)

  diamond makedb --in $FILE_PROTEIN_SEQ --db $FILE_PROTEIN_DB
  diamond blastp --db $FILE_PROTEIN_DB --query $FILE_PROTEIN_QEURY --ultra-sensitive --outfmt 6 \
    --out $FILE_OUTPUT --max-target-seqs 1 --evalue 0.00001 --block-size 20 --tmpdir $FOLDER_TEMP \
    --no-unlink --threads 12 --masking 0 --comp-based-stats 3

• Each query protein, the best match protein in nr was selected based on the highest bitscore and combined.
• BLASTP against best match proteins.
• Calculate e-value against all nr sequences.
  • As the selected proteins from nr was a subset of the large databases, the e-values were smaller due to the smaller database size. The e-values against the entire nr were calculated by multiplying a ratio, which equaled the total length of proteins in nr dividing the total length of the selected nr proteins.

UniProt

Similar to nr.

SwissProt

Search with BLASTP directly.