What Really Counts

Is it the gene or the expression of the gene? Is it the DNA or the RNA? As we progress in understanding cancers we often look for the putative gene. DNA can be read and the presence of absence of the gene is considered dispositive. However, it often is not.

The real is is; does the gene do anything? Namely does it produce an RNA and in turn a protein? In a recent paper Biorxiv, CSHL, we have an interesting result. The authors note:

Disappointing results with genome-based biomarkers have driven calls to look beyond the cancer genome to other possible indicators of cancer-specific vulnerabilities​5​. A major alternative for tumor characterization is the transcriptome. Although RNA is more difficult to maintain in the clinic than DNA, studies have found that gene expression supplies the most significant predictive features for patient prognosis​6​. Previous work on predicting cell viability after compound treatment suggests that expression may be more powerful than DNA features for predicting drug response​7​, but expression-based models are widely treated as undesirable either because they are considered trivial proxies for tissue-type​8​ or because they are seen as uninterpretable​9​. These suggestive studies point to the need  for a comprehensive comparison of expression and genomic molecular features as predictors of cancer vulnerability and a deeper interrogation of the interpretability of expression models. Here we present the first such study across five large datasets of cancer cell viability including both genetic and chemical perturbations. We find that RNA-Seq expression outperforms DNA-derived features in predicting cell viability response in nearly all cases, including many perturbations with known genomic biomarkers. The best results are typically driven by a small number of interpretable expression features. Our findings suggest that both existing and new cancer targets are frequently better identified using RNA-seq gene expression than any combination of other cancer cell properties.

RNA can be fragile and more difficult to assess. We have seen that RNA-seq does provide a means and have seen this in miRNA analyses. Perhaps this is an added methodology worthy of pursuit.