It has long been appreciated that hemostatic systems represent complicated dynamics, involving multiple factors, which work in concert to regulate the balance between coagulation and bleeding in both health and disease. In this issue of Blood, Deguchi et al present evidence to support a novel role for acylcarnitines as anticoagulants.1 These findings were arrived at starting with an untargeted metabolomics approach, which identified 10:1 and 16:1 acylcarnitines as decreased in plasma from patients with venous thrombus and pulmonary embolism (VTE) compared with a control group. A follow-up targeted approach was used as a means of method validation. Importantly, the addition of acylcarnitines to clotting assays demonstrated a direct effect through inhibition of factor Xa. Moreover, the chain length of the acylcarnitines was important, with carbon lengths of over 14 having the highest activity. Finally, mapping studies were carried out to analyze binding sites of acylcarnitines with factor Xa.
Philosophers of science have long made the distinction between the logic of discovery and the logic of justification. At the end of the day, many (although not all) philosophers of science have abandoned the program to define a distinct logic of the process of discovery. In contrast, the logic of justification, the process by which scientific claims are evaluated, continues to mature in a highly meaningful way. In many ways, the advent of “omics”-type approaches to discover new knowledge has rejuvenated issues of the logic of discovery and, in doing so, both generated new capacities and resurrected old problems. Mill’s method of comparing what variables are common to one state (eg, a disease) but different in another (eg, health) and of attributing a potentially causal role to observed differences is essentially the basis of discovery-level untargeted metabolomics.2 Because omics cannot assess all possible variables, one cannot go as far as Mill claimed and assess causality, but only correlation. A more rigorous evaluation of causality requires isolation of a single variable, either through study design or interventional experimentation (with the removal of 1 variable, only 1 variable, and without changing anything else; a high threshold for success indeed). Given both the technical and ethical restrictions of human studies, such is not a possibility; however, Deguchi et al take additional measures to address the correlation/causality issue.
First, additional targeted and quantitative metabolomics were carried out as a means of validating the accuracy and precision of the antecedent and unbiased discovery-based untargeted approach. Second, having accomplished the discovery phase goal of identifying acylcarnitines as candidate molecules for playing a causal role, the investigators tested a prediction deduced from their new acylcarnitine hypothesis; in particular, that acylcarnitines would have a functional effect on coagulation itself. Indeed, using controlled assays of coagulation, it was observed that acylcarnitines had a hitherto unappreciated activity of inhibiting factor Xa-dependent coagulation assays. Thus, both verification of the untargeted analyte screen was carried out and the hypotheticodeductive process was explored, testing predictions from the hypothesis that acylcarnitines play a role in VTE (eg, that acylcarnitines would have activity in coagulation assays). Together, these data provide provocative evidence of a new pathway, which may yield insight into both the biology of coagulation and a potential target for therapeutic intervention.
As a general property of omics-type explorations of natural phenomena, several essential next steps are required. The first is a mindful consideration that traditional statistics and metrics for significance (eg, P values with a .05 cutoff) are not meaningful in the context of many omics-based approaches. By traditional metrics, a type I error will be made in 1 of every 20 studies by chance alone (measuring a single variable). However, omics-type approaches measure hundreds of variables on each specimen, and for every 100 variables, 5 will be significant by chance alone. Statistics has evolved to test the deviation of P values from this predicted normal distribution, giving rise to more stringent metrics of significance (ie, q values)3 ; however, although a low q value indicates likely significance, a higher q value does not rule it out. Moreover, such statistical considerations assume that each variable is independent of each other variable, which is clearly not the case for metabolic pathways, and thus the nature of statistical predictions changes in omics-type studies. A second concern is what is meant by “verification.” A targeted analysis of the same samples on which the untargeted approach was carried out is a form of method verification and tests the accuracy and precision of the untargeted semiquantitative data. However, although this verification addresses whether the observed correlations actually occurred in the cohort studied, it does not verify that such correlations did not happen by chance alone. To evaluate this latter question, a new cohort, distinct from the group analyzed to generate the initial observation, must be analyzed. If the same correlation is observed in a separate cohort, then this will provide much confidence.
The above cautions notwithstanding, the data put forth by Deguchi et al demonstrate a distinct biological basis for how acylcarnitines might play a causal role in VTE pathogenesis. These findings represent a new view of coagulation regulation, involving a class of compounds not previously appreciated to be intricate to this process. The previously ubiquitous and unnecessarily pejorative descriptor of “fishing expedition” has recently been replaced with the somewhat euphemistic label of “hypothesis-generating studies.” In this case, Deguchi et al have caught a very nice fish and generated a provocative and potentially seminal hypothesis.
Conflict-of-interest disclosure: J.C.Z. has a sponsored research agreement with Immucor Inc. that is unrelated to the current studies.
