Query dataset and sequence search

We initially curated the human TIR domain sequences from Uniprot and filtered the hits to include Swiss-Prot reviewed sequences. Further, amongst these, those which followed PROSITE-ProRule annotation: {"type":"entrez-protein","attrs":{"text":"PRU00204","term_id":"1359536640","term_text":"PRU00204"}}PRU00204, were only selected. This PROSITE profile is specific to the pattern of TIR domains’ scaffold that promotes assembly of signaling complexes via protein–protein interactions. Motif search was performed using MEME suite by classic search method and gapped local alignment was performed by the GLAM2 module using protein seqeuences as input structure [26–28]. Such human TIR sequences were used as a query to search for best homologs using BLAST [29] with an e-value of 10^–10. Best representatives of Primates, Odd-toe ungulate, Even-toe ungulate, Carnivore, Placental, Whale and dolphins, Chiropteran, Rodentia, Lagomorpha, and Insectivores were selected for each query. The motif search was extended to this set of proteins (275 proteins).

In order to search the homologues of TRAM and TRIF, we enriched our dataset to include organisms from 22 orders across the Mammalian class. TRAM and TRIF proteins of different organisms from the following orders were included. Monotremata (Ornithorhynchus anatinus), Didelphimorphia (Monodelphis domestica), Dasyuromorphia (Sarcophilus harrisii), Diprotodontia (Phascolarctos cinereus), Cingulata (Dasypus novemcinctus), Proboscidea (Loxodonta africana), Afrosoricida (Echinops telfairi), Tubulidentata (Orycteropus afer afer), Rodentia (Marmota flaviventris), Primates (Pan troglodytes), Eulipotyphla (Condylura cristata), Chiroptera (Pteropus vampyrus), Artiodactyla (Sus scrofa), Cetacea (Balaenoptera acutorostrata scammoni), Perissodactyla (Equus caballus), Carnivora (Felis catus), Lagomorpha (Ochotona princeps), Macroscelidea (Elephantulus edwardii), Scandentia (Tupaia chinensis), Dermoptera (Galeopterus variegatus), Sirenia (Trichechus manatus latirostris), Pholidota (Manis javanica) orders along with two model organisms (Mus musculus, Macaca mulatta) protein and Homo sapiens.

Later, the TRAM and TRIF protein from 25 organisms were used as a query to perform CS-BLAST [30] against the NR_Sept2019 database. The search was done using a python script to include all sequences individually with a very stringent e-value of 10^–10 and up to 5 iterations after which it got saturated. The results from all CS-BLAST searches were combined and using an in-house script the output of CS-BLAST was converted into a tabular format. These results were further filtered using a query coverage filter of more than or equal to 50% and a sequence identity filter of more than or equal to 30% (keeping in mind the Twilight zone of protein sequence alignment) [31]. The list of hits obtained from multiple query search after considering the query coverage and sequence identity cut off for TRIF and TRAM orthologues are shown in Additional Files 14 and 15 respectively. The sequence of reference ID from these hits was retrieved using blastdbcmd module of BLAST version 2.9.0+. To remove the redundancy amongst sequences, CD-HIT was used to cluster the hits with 100% identity cutoff [32]. These hits were further divided into subfamilies based on clustering pattern by constructing sequence similarity network (SSN). These classifications were done using ZEBRA2, based on the CD-HIT clustering approach with a sequence identity threshold of 30% [10]. Based on subfamily clusters and phylogeny, proteins homologous were categorized into TRAM and TRIF family. Here we also used the protein sequence from PDB entries 2M1X and 2M1W, that corresponds to TIR domain region from TRIF and TRAM human protein respectively. We used these TIR sequence along with the full-length TRIF and TRAM orthologs sequence to see where does the TIR sequence clusters in the SSN.