LAUSR

‘But, where’s the beef?’, a colleague and friend asked me at a conference, right after a student from my group had presented our first transcriptional data set on T cells in viral hepatitis. I knew exactly what she meant. As scientists, we are accustomed to expect clear and definitive answers to specific questions that step by step prove or disprove a defined hypothesis developed on rigorous previous data. However, comprehensive and unbiased omics studies, like the analysis of genomewide gene expression in liver tissue with HCV infection by Marchi et al in Gut, are in most instances neither aimed nor suited to deliver easytograsp answers. This is especially true for crosssectional analyses in humans with their inherent heterogeneity in mutable and immutable traits. The conversation mentioned previously happened almost 10 years ago, and studies using omics approaches in human cohorts have become commonplace in the meantime, but most of us still feel not completely at ease assessing the merits of omicsbased work. This challenge is aggravated by the rapid speed with which techniques for omics data generation as well as the computational tools for their analysis are evolving, making one feel always one or two steps below the required level of expertise. Without a doubt, coming generations of scientists will deal with largescale omics data much more naturally, but for now the challenge of evaluating and understanding complex omics studies is a reality for many of us. The transcriptomics study by Marchi et al serves as an excellent example to discuss criteria for identifying those omics studies that are highly informative now and that can be expected to have significant impact down the road. Key aspects are the significance and potential of the analysed material, the quality of the data necessary to reveal important biological differences and the likeliness of the data to be useful for others beyond the analysis presented by the investigators themselves. With regard to the analysed material, Marchi et al performed RNAseq on whole liver tissue from a staggering 195 patients with chronic HCV infection. There can be no doubt that highquality descriptive data from diseased human tissue are indispensable to inform subsequent hypothesisdriven work and for the development of model systems that actually reflect the disease of interest. As liver biopsies are increasingly difficult to procure, the availability of almost 200 tissues available for analysis is a major differentiating strength of this study. Sampling on that scale ensures the power to distil broadly shared differences as well as more subtle traitassociated features in heterogenous human populations. This asset trumps some limitations of the study, such as the lack of singlecell data or healthy liver controls. Singlecell data or use of the latest multiomics approaches, for example, combining RNAseq and ATACseq, would have certainly delivered additional valuable information. However, something new is always around the omics corner, and the continuous wait for the next leap in technologies can stifle progress needed right now. Appreciating data based on the valuable information they contain is more important than lamenting about what could be done with the material now. In the case of HCV hepatitis (and this most likely holds true for most other liver diseases), it is extremely unlikely that we will see a similarly sized biopsy study in the future, assuring that the material studied by Marchi et al is uniquely poised to deliver significant and lasting insights. Having established the potential of the patient cohort and the material studied, the quality of the data and their analysis need to be determined. An initial assessment can be on a purely technical level, checking quality control parameters and the appropriateness of analytical tools. In my opinion, this step is best performed by specialists in the respective omics approaches and should be done during peer review, either by inhouse editors or by including reviewers who are more experts in technology than in disease. At the current moment, this is not the standard process and it is not rare that one can spot obvious technical flaws in published manuscripts, even as somebody with a moderate understanding of the technical underpinnings of omics studies. The second level of quality control is to check for biological plausibility. In the present study, the first finding is that unsupervised clustering of the samples rather stringently differentiates samples with and without cirrhosis, which is reassuring, given the broad and complex transcriptional changes known to associate with fibrosis. Beyond this dominant layer of differential gene expression, biological sex and IL28B/IFNL4 genotype had the most notable impact on the intrahepatic transcriptional landscape. Linking these traits to specific cellular processes required a more advanced analytical approach called weighted gene coexpression network analysis (WGCNA). This method goes beyond differential gene expression by identifying modules of coregulated genes, with coregulation of a group of genes identifying molecular programmes likely linked to biological processes. Using this approach, Marchi et al found strong evidence that women display stronger interferonresponse signalling, while men have increased expression of T cellrelated genes. This is highly significant as women have better HCV infection outcomes, both regarding spontaneous control of viremia as well as disease progression. The findings demonstrate that even whole liver tissue transcriptome data can allow surprisingly granular and biologically relevant findings that can be explored subsequently by more targeted approaches, be they experimental or analytical. The final important aspect of valuable omics studies is the potential to be a resource for future investigations. Indeed, Marchi et al themselves used existing data sets to strengthen their analysis and conclusions. Their own data set should also become an important resource, as it is extremely likely that singlecell multiomics studies will have to rely on much smaller numbers of liver tissues. Novel approaches to analyse and integrate omics data sets across different platforms and experimental batches are developed Division of Gastroenterology, Massachusetts General Hospital, Boston, Massachusetts, USA Harvard Medical School, Boston, Massachusetts, USA