Omics-Based Interaction Framework – a systems model to reveal molecular drivers of synergy

Bioactive molecule library screening strategies may empirically identify effective combination therapies. However, without a systems theory to interrogate synergistic responses, the molecular mechanisms underlying favorable drug-drug interactions remain unclear, precluding rational design of combination therapies. Here, we introduce Omics-Based Interaction Framework (OBIF) to reveal molecular drivers of synergy through integration of statistical and biological interactions in supra-additive biological responses. OBIF performs full factorial analysis of feature expression data from single vs. dual factor exposures to identify molecular clusters that reveal synergy-mediating pathways, functions and regulators. As a practical demonstration, OBIF analyzed a therapeutic dyad of immunostimulatory small molecules that induces synergistic protection against influenza A pneumonia. OBIF analysis of transcriptomic and proteomic data identified biologically relevant, unanticipated cooperation between RelA and cJun that we subsequently confirmed to be required for the synergistic antiviral protection. To demonstrate generalizability, OBIF was applied to data from a diverse array of Omics platforms and experimental conditions, successfully identifying the molecular clusters driving their synergistic responses. Hence, OBIF is a phenotype-driven systems model that supports multiplatform exploration of synergy mechanisms.

Interaction Introduction 5 the indicated inhaled treatments, we observed little increase in survival after the individual treatments compared to sham-treated control mice, whereas mice treated with the Pam2-ODN combination 6 Differentially expressed molecules reveal synergy-specific pathways To investigate the mechanisms underlying Pam2-ODN synergy, we used OBIF to re-analyze previously 7 V and VI; F B dominant, VII and VIII). Principal component analysis ( Figure 2D) demonstrates that 153 concordant EPs (I and II) were the most abundant in our dataset, followed by Pam2-dominant profiles 154 (V and VI). This abundance of EPs I and II better emphasizes the cooperative effects of both factors 155 than does conventional DEM clustering alone. In particular, the contribution of ODN to the synergistic 156 combination might otherwise be overlooked by DEM analysis, as it induces enrichment of far fewer 8 II) are influenced by at least one multi-factor effect, while all features in discordant profiles (III and IV) also revealed that 67% (2116/3138) of Pam2-ODN DEMs are driven by the interaction effect of Pam2 182 and ODN as interacting DEMs (iDEMs) ( Figure 3C). Thus, OBIF reconciled the biological interactions 183 from EPs with the statistical interactions from multi-factor effects of Pam2-ODN. where CI is the absolute ratio of the log2 fold change of Pam-ODN-induced DEMs (F AB ) and the 205 additivity threshold of Pam2 (F A ) and ODN (F B ), allowing identification of both antagonistic (CI < 1) or 206 synergistic (CI > 1) features ( Figure 3F). A log2 transformation of the CI then yields an interaction score 207 (IS) that quantifies the effect size of non-additive expression relative to the additivity threshold, and can 208 be applied to both antagonistic (IS < 0) and synergistic (IS > 0) iDEMs ( Figure 3G). This allows more 209 focused enrichment analysis, in this case supporting NF-κB/RelA and AP-1/cJun as key transcriptional

228
To differentiate transcriptional cooperation from coincidental transcription factor activation after Pam2-

229
ODN treatment, we assessed the Pam2-ODN-induced nuclear co-translocation of NF-κB and AP-1 10 complexes in the presence or absence of NF-κB inhibitor IMD-0354 (IMD). As expected, pre-treatment 231 with IMD alone reduced the Rd Value and percentage of translocated cells for RelA and p50 without 232 significantly modifying the percentage of translocation for cJun and cFos. However, NF-κB inhibition 233 with IMD also unexpectedly reduced the Pam2-ODN-induced similarity score shifts and nuclear 234 translocation of AP-1 subunits, particularly of cFos ( Figure 4E). This indicates that NF-κB inhibition 235 impaired Pam2-ODN-induced AP-1 nuclear translocation, confirming the cooperative regulation of 236 these two non-overlapping signaling pathways. Representative images shown in Figure    datasets to deduce the mechanisms mediating the synergy. This is important because, although this 278 lack of mechanistic understanding does not limit the utility of the current combination, it precludes 279 development of next generation interventions that more precisely (perhaps, more efficaciously) target 280 the synergy-driving pathways with fewer off-target (potentially toxic) effects. In contrast to models that 281 predict possible synergy, OBIF was developed with the explicit intent to investigate established 12 synergistic events. As such, it is inherently a phenotype-driven model that performs full factorial 283 analysis on feature expression data from single vs. dual factor exposures to identify molecular clusters 284 that reveal synergy-mediating pathways, functions and regulators.

Reagents and Tools
The subscripts denote the condition of the samples: exposed to neither factor (0,0), exposed to factor A 385 alone (1,0), exposed to factor B alone (0,1) or exposed to both factors A and B (1,1). The superscripts 386 represent the sample replicates from 1 to i within each of the four conditions.

388
To improve detection of interaction effects, OBIF allows sequential transformation of an unscaled To evaluate significant interaction terms between factors at the whole "-ome" level, OBIF performs a 401 multiple linear regression across the expression values in a dataset: where the interaction analysis of the Omics expression levels (E O ) is equivalent to a two-way ANOVA 404 analysis where the intercept is referenced to the control samples (0)

442
The standard errors calculated from the interaction term use a significance threshold (p-value < 0.05) to 443 determine significant interaction effects. The spatial relationship between the transcription factors and nuclear images was measured using the 526 'Similarity' feature in the IDEAS software to quantitate the mean similarity score in the cell populations