C. elegans protein interaction network analysis probes RNAi validated pro-longevity effect of nhr-6, a human homolog of tumor suppressor Nr4a1

Protein-protein interaction (PPI) studies are gaining momentum these days due to the plethora of various high-throughput experimental methods available for detecting PPIs. Proteins create complexes and networks by functioning in harmony with other proteins and here in silico network biology hold the promise to reveal new functionality of genes as it is very difficult and laborious to carry out experimental high-throughput genetic screens in living organisms. We demonstrate this approach by computationally screening C. elegans conserved homologs of already reported human tumor suppressor and aging associated genes. We select by this nhr-6, vab-3 and gst-23 as predicted longevity genes for RNAi screen. The RNAi results demonstrated the pro-longevity effect of these genes. Nuclear hormone receptor nhr-6 RNAi inhibition resulted in a C. elegans phenotype of 23.46% lifespan reduction. Moreover, we show that nhr-6 regulates oxidative stress resistance in worms and does not affect the feeding behavior of worms. These findings imply the potential of nhr-6 as a common therapeutic target for aging and cancer ailments, stressing the power of in silico PPI network analysis coupled with RNAi screens to describe gene function.

in lifespan reduction. On randomly picking up the 1727 proteins (reducing the 290 high network importance proteins from the total network proteins) with no network importance, the probability of picking up LRGs will be only 3.53%. Conclusively, picking the high network importance gene highly increased the chances of success during the experiments.
Although these were a random expectation and following points should be considered before any biological interpretation about a priori probabilities -The network is limited to aging genes and TSGs, although each gene in the network may have more binding partners (outside our network). This can subsequently impact the classification of network importance genes.
-New candidate gene for lifespan reduction might have or not have direct interactions with the known LRGs. This might be due to indirect regulation or unknown PPIs.

Prospects of randomly picked C. elegans gene from cluster 8, 1, 4 or 6 to reduce lifespan upon inhibition
To calculate these, we build a new cluster (cluster 9: derived from cluster 1, 3 and 4) in the network consisting the experimentally verified lifespan reduction genes coded proteins as annotated in GenAge database (  potential to alter lifespan can also be different. Indeed, it also cannot be ignored that many connections in the network might be missing therefore applying GBA to predict the lifespan reduction potential may be oversighted. To overcome this, we measured the semantic similarity between the individual proteins of all the clusters and LRGs (cluster 9). We annotated the GO of all the proteins from different clusters and further identified the GO categories overrepresented for cluster 9 and then accessed the semantic similarities score (for molecular function) using recently published GOGO algorithm (Zhao and Wang, 2018). It should be noted that many proteins could not be annotated with GO class and many proteins could not be scored because of no similarity of GO terms between the individual protein and cluster 9. Although the annotated proteins are enough to access the lifespan reduction possibilities. Among the top three genes that we have discussed in the manuscript, only nhr-6 was accessed by this as gst-23 could not be annotated by GO-term and vab-3 lacks semantic similarity with LGRs.
We identified the semantic similarity of nhr-6 GO terms with LRGs as 0.9. This indicated the higher probability of lifespan reduction on removing this protein.
Here we compare this score with the proteins from the other clusters Cluster 8 consist of protein annotated as TGSs coded proteins. The maximum score we noticed was 0.42 ( Fig.   S2) which indicates their functional difference with LRGs.
Cluster 1 consist of proteins annotated to be involved in aging. The maximum score we noticed was 0.9 for xbp-1 (Fig. S3). Xbp-1 is annotated as aging gene in the GenAge database but its influence on longevity is still unannotated (http://genomics.senescence.info/genes/search.php?search=xbp-1). Based on the scoring, we believe that this gene might also be involved in lifespan reduction. Supporting this, it has been reported that the overexpression of xbp-1 in muscle cells reduces the lifespan by 25% (https://www.ncbi.nlm.nih.gov/pubmed/23791175). Beside this, other proteines in the cluster scored < 0.6.
We could only score cep-1 in cluster 4 which showed 0.38 semantic similarity score with LRGs. Comparatively with nhr-6 the score was very less to claim any lifespan reduction involvement of cep-1.
In cluster 6 proteins, we found sex-1 with the 0.9 score like nhr-6 which is also the member of same cluster ( Fig. S4). Among other proteins, let-418 was scored 0.81. However, let-418 mutants showed the increased lifespan attributed to its regulatory functions (De Vaux et al., 2013). For the other proteins in this cluster the score was < 0.4. Evidently, plx-1, one of the bottom proteins in the cluster 6 based on its topological importance, was scored only 0.08. Figure S1. Protein-protein interaction network of aging and human tumor supressor gene orthologs of C.
elegans. Network is organized into nine clusters. Eight clusters are same as shown in Fig. 1c. Cluster 9 is added which is derived from Cluster 1, 3 and 4 and contains lifespan reduction genes (LRGs). The annotation of the nodes can be accessed from the XGMML (extensible graph markup and modelling language) network Supplementary File 3.