04.12.results_giant.md

Replication of gene associations using tissue-specific gene networks from GIANT

We sought to systematically analyze discrepant scores to assess whether associations were replicated in other datasets besides GTEx. This is challenging and prone to bias because linear-only correlation coefficients are usually used in gene co-expression analyses. Therefore, we used 144 tissue-specific gene networks from the Genome-wide Analysis of gene Networks in Tissues (GIANT) project [@pmcid:PMC4828725; @url:https://hb.flatironinstitute.org], where nodes represent genes and each edge a functional relationship weighted with a probability of interaction between two genes (see Methods). The version of GIANT used in this study did not include GTEx samples [@url:https://hb.flatironinstitute.org/data], making it an ideal case for replication. These networks were built from expression and different interaction measurements, including protein-interaction, transcription factor regulation, chemical/genetic perturbations and microRNA target profiles from the Molecular Signatures Database (MSigDB [@pmid:16199517]). We reasoned that statistically significant and highly-ranked gene pairs using three different coefficients in a single tissue (whole blood in GTEx, Figure @fig:upsetplot_coefs) that represented real patterns should often replicate in a corresponding tissue or related cell lineage using the multi-cell type functional interaction networks in GIANT. In addition to predicting a network with interactions for a pair of genes, the GIANT web application can also automatically detect a relevant tissue or cell type where genes are predicted to be specifically expressed (the approach uses a machine learning method introduced in [@doi:10.1101/gr.155697.113] and described in Methods).

As an example of our evaluation procedure, we obtained the networks in blood and the automatically-predicted cell type for gene pairs RASSF2 - CYTIP (strong nonlinear pattern and CCC high, Figure @{fig:giant_gene_pairs}a) and MYOZ1 - TNNI2 (weak linear pattern and Pearson high, Figure @{fig:giant_gene_pairs}b). In addition to the gene pair, the networks include other genes connected according to their probability of interaction (up to 15 additional genes are shown), which allows estimating whether genes are part of the same tissue-specific biological process. Two large black nodes in each network's top-left and bottom-right corners represent our gene pairs. A green edge means a close-to-zero probability of interaction, whereas a red edge represents a strong predicted relationship between the two genes. In this example, genes RASSF2 and CYTIP (Figure @{fig:giant_gene_pairs}a), with a high CCC value ($c=0.20$, above the 73th percentile) and low Pearson and Spearman ($p=0.16$ and $s=0.11$, below the 38th and 17th percentiles, respectively), were both strongly connected to the blood network, with interaction scores of at least 0.69 and an average of 0.77 and 0.85, respectively (Supplementary Table @tbl:giant:weights). The autodetected cell type for this pair was leukocytes, and interaction scores were similar to the blood network (Supplementary Table @tbl:giant:weights). However, genes MYOZ1 and TNNI2, with a very high Pearson value ($p=0.97$), moderate Spearman ($s=0.28$) and very low CCC ($c=0.03$), were predicted to belong to much less cohesive networks (Figure @{fig:giant_gene_pairs}b), with average interaction scores of 0.17 and 0.22 with the rest of the genes, respectively. Additionally, the autodetected cell type (skeletal muscle) is not related to blood or one of its cell lineages. These preliminary results suggested that CCC might be capturing blood-specific patterns missed by the other coefficients.

{#fig:giant_gene_pairs width="100%"}

We next performed a systematic evaluation using the top 100 discrepant gene pairs between CCC and the other two coefficients. For each gene pair prioritized in GTEx (whole blood), we autodetected a relevant cell type using GIANT to assess whether genes were predicted to be specifically expressed in a blood-relevant cell lineage. For this, we used the top five most commonly autodetected cell types for each coefficient and assessed connectivity in the resulting networks (see Methods). The top 5 predicted cell types for gene pairs highly ranked by CCC and not by the rest were all blood-specific (Figure @{fig:giant_gene_pairs}c, top left), including macrophage, leukocyte, natural killer cell, blood and mononuclear phagocyte. The average probability of interaction between genes in these CCC-ranked networks was significantly higher than the other coefficients (Figure @{fig:giant_gene_pairs}c, top right), with all medians larger than 67% and first quartiles above 41% across predicted cell types. In contrast, most Pearson's gene pairs were predicted to be specific to tissues unrelated to blood (Figure @{fig:giant_gene_pairs}c, bottom left), with skeletal muscle being the most commonly predicted tissue. The interaction probabilities in these Pearson-ranked networks were also generally lower than in CCC, except for blood-specific gene pairs (Figure @{fig:giant_gene_pairs}c, bottom right).

The associations exclusively detected by CCC in whole blood from GTEx were more strongly replicated in these independent networks that incorporated multiple data modalities. CCC-ranked gene pairs not only had higher probabilities of belonging to the same biological process but were also predicted to be specifically expressed in blood cell lineages. Conversely, most Pearson-ranked gene pairs were not predicted to be blood-specific, and their interaction probabilities were relatively much lower. This lack of replication in GIANT suggests that top Pearson-exclusive-ranked gene pairs in GTEx might be driven mainly by outliers, which is consistent with our earlier observations of outlier-driven associations (Figure @{fig:upsetplot_coefs}b).