It is widely recognized that central nervous system (CNS) disorders are major contributors to global disease burden. Yet although progress in basic neuroscience is accelerating at a rapid pace, there has been limited translation of our understanding of brain function—and dysfunction—into novel therapies and treatments for CNS disorders. This could in part be the case because investigators still face a fundamental unanswered question: what constitutes a 'good' animal or cellular model of neurological disease?

Experimental investigation into the pathogenesis of CNS disorders is an essential step toward ultimately developing new and effective treatments for patients. The translational potential of such research, however, hinges on the utilization of model systems that are etiologically relevant and that recapitulate some essential feature(s) of the human disease of interest. Animal and cellular models of CNS disorders have historically been judged by their respective validation in three core domains. Construct validity refers to the extent to which the disease etiology in the model accurately reflects that of humans; face validity refers to the degree to which phenotypic observations of the model system resemble those of the human condition; and predictive validity describes the usefulness of a model as a means of informing the translation of treatments from the model into therapies for humans.

The findings made in these model systems might not translate into efficacious treatments for human patients for many practical reasons. The pathway of interest may not be conserved between species. Blood-brain barrier permeability, target engagement, safety, dosing and drug metabolism might differ between small animal or cellular models and humans. Conceptually, however, a remaining major concern, as mentioned earlier, centers on the extent to which these systems represent 'good' models of human disease in the first place.

Investigators in the preclinical neuroscience research communities are acutely aware of the strengths and weaknesses of different animal and cellular models, and they have initiated efforts to disseminate this information. Within the Alzheimer's disease community, for example, researchers have meticulously described more than 100 different transgenic mouse models that may be used in preclinical neurodegenerative disease research—delineating the constructs used to generate the models as well as the specific time course of pathological and cognitive phenotypic development—in the ALZFORUM Research Models database. However, although this type of detailed information may be available in specific communities, it is often not included in the scientific publications that report new findings using these models. The omission of this information makes it difficult for researchers in other fields to judge the translational potential of such findings because they lack full knowledge of the strengths and weaknesses of the chosen model systems. Yet the issue of model system utility has received an encouraging amount of attention over the past several years, for example in a Perspective by Tobias Kaiser and Guoping Feng in this issue of Nature Medicine (page 979).

For this reason, moving forward, Nature Medicine will work with authors of accepted preclinical neuroscience research articles to ensure that a brief introduction and rationale for the use of a particular model system is included in the manuscript. We hope that a detailed description of model validity in the relevant domains mentioned above will help to explain to a broad biomedical audience why a given cellular or animal model was chosen as a hypothesis-generating or testing system. Furthermore, we feel that a discussion of the reasons that a particular set of molecular, anatomical, electrophysiological and behavioral phenotypes were chosen for analysis, as well as how these might relate to the human disease condition, will increase the transparency of our preclinical CNS disease research. An additional consideration is whether these measures are likely to be used as clinical endpoints for human trials of new therapies that may be designed based on the findings of the present study.

With regard to animal modeling of neuropsychiatric and neurodevelopmental disorders, we would also encourage investigators to discuss their findings within the framework of the NIMH's Research Domain Criteria (RDoC) initiative in addition to the diagnostic criteria outlined in the latest edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5). As a research framework, the RDoC initiative has called for a shift toward studying disease with a focus on basic dimensions of mental function that cut across current diagnostic boundaries. From this perspective, investigations into cellular, circuit and behavioral endophenotypes of a single model may lead to a better understanding of the pathophysiology of multiple disorders.

We will also take these considerations into account during earlier stages of the editorial process, for example when determining whether or not to send a manuscript out for external review and when providing guidance to authors on which referee requests for additional experimental data are of highest priority. We hope that these efforts will promote a more transparent discussion of CNS disease modeling among authors, referees and readers. Open discussion will aid interpretation of results based on animal and cellular models with implications for the pathogenesis and treatment of CNS disorders. As always, we welcome your feedback and suggestions at medicine@us.nature.com.