Mapping Pathways to Drug Mechanism of Actions

A hallmark of targeted molecular therapies is over-expression of the drug's molecular target. Strong support for this behavior can be found within the NCI60 screen as seen by the positive correlations between gene expressions of the proteasome and heat shock proteins to Velcade® and Geldanamycin®, respectively (17). Extending these cytotoxicity-gene expression correlations to pathways is an attempt to establish a pathway-centric perspective to a drug's MOA.

Instead of examining each individual drug-gene correlation, correlations between the GI50 SOM clades (clusters of drugs with similar GI50 profiles) and pathways are evaluated as a more general approach. For each pathway, correlations with GI50 clades are compared between genes "on a pathway" with genes "off a pathway" and the Kruskal-Wallis H statistic is calculated (57,58). Each pathway in KEGG, BioCarta, and each GO term is then mapped onto the GI50 SOM, where each clade has an H-score, representing the strength of correlation between the pathway and the compounds in that clade. The most significantly and specifically correlated pathways are proposed as the most likely targets of the drugs within a clade. Collectively, signal transduction pathways are among the least cohesive pathways; their correlations with the GI50 SOM regions are generally diffuse, lacking assignment to any one clade or region in GI5o response space. Conversely, this may also imply that it will be difficult to find drugs that specifically target signaling pathways.

One of the primary goals of the drug-pathway analysis is to find and interpret drug targets and MOA. The biological pathways that are potentially perturbed by the drugs in the nine SOM response regions can be postulated. Conversely, each pathway may be associated with one or more response regions. We have previously established the general MOA of the agents in some of these SOM regions: mitosis (M), membrane function and oxidative stress (N), nucleic acid metabolism (S), and metabolic stress and cell survival (Q), oxidative metabolism (R), and kinases/phosphatases and oxidative stress (P), via other methods (17,18,19).

The pathway mapping results provide additional support for the annotation of some of the SOM regions: For example, the pathways mitotic checkpoint, cytokinesis, kine-tochore, and cell cycle are associated with the M-region; the mitochondrial inner membrane, response to oxidative stress, and oxidoreductase activity are associated with the N-region; the granzyme A mediated apoptosis pathway, the DNA topological change, and DNA topoisomerase (ATP-hydrolyzing) activity, are associated with the S-region; the pathways relating to glutamate metabolism, xenobiotic metabolism, cysteine metabolism, and glutathione biosynthesis are associated with the Q-region; the pathways of fatty acid metabolism and oxidative phosphorylation and NADH dehydrogenase (ubiquinone) activity, NADH dehydrogenase activity, oxidoreductase activity, and mitochondrion are associated with the R-region; and the pathways in signaling of hepatocyte growth factor receptor, Erk1/Erk2 MAPK signaling pathway, ATM signaling pathway, FAS signaling pathway (CD95), and oxidoreductase activity, cell-cell signaling and DNA damage response (signal transduction resulting in induction of apoptosis) are associated with the P-region.

The additional pathways that are associated with each SOM region through this global pathway analysis provide valuable information and new insights into the MOA for similarly clustered drug molecules. Of interest are the associations of apoptosis with the M-region; cell adhesion and immune response signaling pathways with regions N and P; transport with the N-region; hypoxia and angiogenesis with the P-region; DNA replication, regulation of DNA repair, and translation with the Q-region; and cytoskeleton with the R-region.

Finally, the MOA of the agents clustered in the three regions—F, J and V—can be postulated by examining the pathways associated with these regions: For example, the amino acid metabolism pathways and the Wnt signaling pathway map to the F-region, cell cycle and DNA damage related pathways map to the J-region, and urea cycle and metabolism of amino groups and pyruvate metabolism map to the V-region, which seems for the latter to share pathways with its neighboring regions. In fact, many pathways are shared among different SOM regions, and conversely each region is usually associated with multiple pathways. This is expected because any one biological process can be perturbed by many drugs but to different degrees.

The agents that can most effectively disrupt a process can now be found by looking at the most significantly correlated sets. Moreover, each SOM region contains the GI50 profiles of thousands of compounds; therefore, it is not surprising that multiple processes, even though usually related, are associated with these compounds. To gain more specific information on the MOA of one compound or a small cluster of related compounds, detailed drug-pathway analysis, as described earlier, is required. The region-wide analysis of biological pathways and drug response, however, provides a global view of biological activities or features shared by large groups of compounds.

As discussed previously, pathways responsible for vital cellular processes or processes that are related to growth or proliferation, specifically in cancer cells, such as those engaged in genetic information processing, cell cycle, energy metabolism, and nucleotide metabolism, are found to have significantly more coherent gene expressions than most signaling and regular metabolic pathways. The cohesive pathways are also found here, in general, to have stronger pathway-drug correlations than non-cohesive pathways because the high level of gene co-expression in cohesive pathways makes it more likely for genes in the pathway to have similar correlation patterns with, or act coherently toward, a drug.

This may imply that cohesive pathways are easier to target, because many drugs seem to be able to significantly disrupt these pathways. Conversely, the correlations of the least cohesive pathways with the GI50 SOM regions are generally diffuse and not strong or specific to any one clade or region. This may be an indication that it will be hard to find drugs that can target non-cohesive pathways or the relationship between drugs, and these pathways are not reflected or easily interpretable by gene-drug correlations.

Therefore, instead of looking at non-cohesive pathways that do not correlate significantly with any drugs, it may be more interesting to examine those non-cohesive pathways that can act coherently toward certain drugs, that is, how correlation or interaction with drugs changes their intrinsic cohesiveness. Taking this one step further, in addition to looking at "drug-coupled" pathway cohesiveness through correlation, insight may be obtained by examining "drug-exposed" pathway cohesiveness, that is, to analyze and compare gene expression cohesiveness within a pathway prior to and after drug exposure.

The number of pathways significantly correlated with each GI50 SOM clade, on the other hand, represents the number of biological processes the drug agents in the clade are potentially perturbing. This number can be used as an indicator of the level of target specificity or promiscuity of these drugs. High pathway correlation promiscuity is indicated for some drugs. Although this may seem undesirable because of multiple targets and thus the potential of detrimental side effects implicated for the drug, this may represent cases where assaulting a single target by the drug can cause multiple intracellular effects, as reflected by correlations with multiple pathways.

On the other hand, this can be deemed to be a desirable property of the drug, because it presents the potential of overcoming the insufficiency of single target inhibition caused by the inherent ability of heterogeneous tumor populations to activate alternative or redundant pathways (59). Based on this premise, drugs with many significantly correlated pathways can be advanced for further investigation.

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook


Post a comment