Conducting a research project

It is important to be familiar with the existing literature, so as to avoid returning to a scientific question that has already been addressed and solved by others. It can be perfectly appropriate to choose to reproduce work that has already been done, but this should be a deliberate choice.

Several articles discuss this issue:

How to increase value and reduce waste when research priorities are set, Chalmers et al., 2014

Value of information: a tool to improve research prioritization and reduce waste, Minelli et Baio, 2015

A recent publication has examined the impact on the advancement of scientific knowledge of the way in which scientific hypotheses are developed and in some cases falsified:

How failure to falsify in high-volume science contributes to the replication crisis, Rajtmajer SM et al., 2022

The NIH guidelines will help you to improve the quality and reproducibility of your proposals, and to meet project evaluation criteria as effectively as possible: Guidance: rigor and reproducibility in grant applications.

The Victoria University website offers general tips on how to respond to a call for tender or a research contract: Preparing a tender or research contract.

The NIH and Hopkins Hospital offer step-by-step methods for responding to calls for research:

How to write an NIH proposal

Guidelines for writing and reviewing a research grant proposal

The content and design of clinical trial protocols are standardized and are required to follow a number of recommendations. For help, you can refer to the following documents, which complement one another.
The WHO has issued recommendations on writing research protocols: Recommended format for a research protocol.

A downloadable guide provides practical information on the characteristics of a valid health research project, the conduct of research, the reporting of research findings and ethical principles: A practical guide for health researchers.

The SPIRIT recommendations for the minimum content of a clinical trial protocol has been approved as an international standard to guide the drafting of clinical trial protocols: Guidance for clinical trial protocols. The electronic version called SEPTRE (SPIRIT Electronic Protocol Tool & Resource) enables you to create your protocol online via an intuitive web interface, and to register it with clinical trial registries: Create, register, and manage trial protocols with SEPTRE.

«Reporting guidelines»: it is important to be familiar with these guidelines, so as to take them into account from the outset of the project, rather than discovering them at the time of publication (in some cases the guidelines have never been collected in the first place). Reporting guidelines can therefore also be considered as «project management guidelines».

On the EQUATOR website, you will find the main recommendations relating to what should be reported in publications, which are elements that many scientific journals require. These recommendations concern, for example, clinical trials, the development of prognostic or diagnostic tools, and pre-clinical studies involving animals. In most cases, there is both a brief version of the recommendations, sometimes in the form of an item list, and a more complete version (elaboration and explanation) which precisely describes the items selected and provides a detailed description of the content of each: enhancing the quality and transparency of research.

Choosing the best possible design for your research, before you start, is a crucial point. Certain tools can help you to design your experiment, either by giving you access to methods for avoiding certain experimental limits, or by giving you access to protocols for work carried out by other teams. To make your research as reproducible as possible, it is important to clarify certain design features of your experiment beforehand. One of these concerns the power analysis needed to estimate the appropriate sample size for your experiment. Others help to implement measures to reduce the risk of subjective bias, such as blinding and randomisation. Asking questions about randomisation, blinding, sample size and potential bias in preclinical studies is essential: A call for transparent reporting to optimize the predictive value of preclinical research, Landis et al., 2012.

There are, for example, tools to improve the design of animal studies proposed by the National Centre for the Replacement, Refinement & Reduction of Animals Research (UK) which can help design robust experiments more likely to produce reliable and reproducible results, among which the Experimental design assistant.

A YouTube presentation is available, and an article in Plos Biology: The experimental design assistant, Percie du Sert et al., 2017.

Protocols.io is an open-access repository where you can find experimental protocols designed by others, so as to enhance the reproducibility of your research. You can also share your protocols to increase the visibility of your own methods.

To raise your awareness towards cognitive bias: Evidence of experimental bias in the life sciences: why we need blind data recording, Holman et al., 2015.

Several publications stress the importance of reproducibility issues in research today, and suggest ways of addressing them at different stages in the research:

A manifesto for reproducible science, Munafo et al., 2017

What does research reproducibility mean?, Goodman et al., 2016

Reproducibility in science. Improving the standard for basic and preclinical research, Begley et Ioannidis, 2015

A report by the American Academy of Sciences, Engineering and Medicine investigates the many aspects of animal research that can contribute to non-reproducible results. The report provides perspectives on (i) how to improve the planning, design and execution of experiments, (ii) the value of reporting all methodological details, and (iii) the value of establishing harmonized reporting principles for the care and use of animals in animal research trials : “Reproducibility issues in research with animals and animal models: workshop in brief”.

In particular, this report reviews the harmonization principles proposed by the International Council for Laboratory Animal Science (ICLAS).

Research involving human subjects: national and international law closely regulates the conduct of experiments on human subjects, but compliance with these rules alone cannot guarantee the acceptability of a research project: ethical questions must be asked before any research project is implemented. It is therefore essential that all research projects involving human beings receive the approval of a research ethics committee (Comité d’éthique de la recherche, CER), before they are implemented. In France, these committees are the CPP (Comité de Protection des Personnes), governed by the Public Health Code, for RIPH projects (Research Involving the Human Person), and the CER (Comité d’Ethique de la Recherche), not governed by law, for non-RIPH research. See, for example: Health Research Qualification guide.

In all instances (RIPH and non-RIPH) the handling of research data is required to comply with the RGPD and the French Data Protection Act (loi informatique et libertés).

The use of animals for scientific purposes: a guide to the ethical evaluation of projects involving the use of animals for scientific purposes has been published by GRICE (Groupe de Réflexion Interprofessionnel sur les Comités d’Ethique), a working group belonging to GIRCOR (Groupe Interprofessionnel de Réflexion et de Communication sur la Recherche) following the recommendations of CNREEA (Comité National de Réflexion Ethique sur l’Expérimentation Animale) and at the request of the French Ministry of Higher Education and Research.

For all matters relating to the “3Rs” principle (replacement, reduction, refinement in animal experiments), a scientific interest group involving Inserm in particular, the FC3R has recently been set up, with the mission of providing training in the 3Rs for the design of projects in line with the 3Rs, of finding funding for these projects, of disseminating relevant information and of issuing calls for FC3R projects.

The acknowledgement of contributions by staff involved in the project should be defined before the start of the research project. You need to specify the conditions of participation of each team and team member in the conduct of the project, to discuss with all participants their expected contributions and the acknowledgement of these contributions, and to conjointly monitor the evolution of these elements throughout the project (order of signatories, etc.). An INSERM document outlines best practices for ensuring the right way to acknowledge the authorship of publications.

A large number of websites will help you complete your knowledge of the regulatory requirements related to your research. These include:

Principles governing research involving human subjects

The French law on research involving humans, known as the Jardé law

The use of animals for scientific purposes

The CNRS research guide is extremely comprehensive. It also provides information on resources that can help to draw up a data management plan, now required by many funding sources and an integral part of research.

Useful information on the RGPD can be found on the CNIL website.

According to the WHO, “registration of all interventional clinical trials is a scientific, ethical and moral responsibility” for all scientists.

Since 2005, it has been mandatory for all clinical trials to be registered on a registry such as Clinical Trials.gov or EuDRACT. This registration is required to take place before beginning the trial, and any modification of the protocol (e.g. of the primary endpoint) should lead to a modification of this registration. No publication is possible without prior registration.

Registration of research protocols is becoming increasingly common in many areas other than clinical trials. In these cases, the term “pre-registration” is used. One article discusses the value of registering preclinical studies: Should preclinical studies be registered?, Kimmelman et Anderson, 2012. Similarly, the following article provides practical details on protocol pre-registration: The preregistration revolution, Nosek et al., 2018.

For example, you can pre-register your protocol on the Open Science Framework (OSF). OSF is developed and hosted by the Center for Open Science (COS), a US-based non-profit organization supported by US federal agencies, private foundations and commercial entities. OSF enables downloading of your project protocol and invites all contributors to work on it with you. You can share it openly so that everyone can see what you are working on. If you are worried about someone “stealing” your idea, you can set up an embargo. In this case, you keep the protocol confidential and reserved for internal use, for a given period, after which it will be made public.

Similarly, PreclinicalTrials is an international online registry for preclinical animal study protocols.
There is also a registry for animal studies managed by the German centre for the protection of laboratory animals (Bf3R). This registry makes it possible to register studies without making them immediately accessible, and to obtain a DOI (Digital Object Identifyer) to identify them.

A model that is gaining ground is the «registered report», in which, on the basis of the protocol, the article is approved in principle before data collection starts. Then, if the analysis has been carried out in compliance with the protocol, the article will be published whatever the result, as described in:

Protocol transparency is vital for registered reports, Chambers et Mellor, 2018

The past, present and future of registered reports, Chambers et Tzavella, 2022

You can also publish the protocol of your work a posteriori in the following journals :

Bio-protocol

STAR protocols

All research projects should be exhaustively documented in a laboratory logbook (electronic or, failing that, paper). Staff from Inserm-affiliated units have access to a secure electronic laboratory logbook solution (CLÉ) called LabGuru. On the Inserm website, you will find several sections on how to set up an electronic laboratory logbook.

The lab logbook is not appropriate for storing the personal data of study participants. If the storage of data requires the entry of individual data, this data must first be pseudonymized (coded), and in all events it can only be released after it has been anonymized (i.e. rendered non-identifiable or, given its level of aggregation, not enabling the identification of the individuals concerned). You thus need to exercise caution on the type of information recorded in your lab logbook.

Statistical analysis can involve a substantial series of choices, which have to be transparently documented in advance in a statistical analysis plan.
Statistical analysis plans are crucial in clinical research, but they should be established for any research project.

Various guidelines are available:

Statistical principles for clinical trials

Guidelines for the content of statistical analysis plans in clinical trials

Quantitative analysis documents

DEBATE-statistical analysis plans for observational studies, Hiemstra et al., 2019

Plos recommendations on how to report statistical analyses

The elaboration of a data management plan (DMP) is mandatory for all ANR-funded projects in France from 2019. On the ANR website, you will find a comprehensive overview of recommendations, practical advice and resources.

Interpreting data means knowing how to avoid the main statistical pitfalls of data analysis. For example you can consult :

Ten simple rules for effective statistical practice, Kass et al., 2016

Science forum: ten common statistical mistakes to watch out for when writing or reviewing a manuscript, Makin et Orban de Xivry, 2019

Sharing your analysis codes boosts confidence in the results, especially when these analyses require complex processing. To make the method used more transparent and reproducible (both for yourself and for others), you can use the following programming tools:

Git, a free, open, distributed version control system that enables you to record your process as you generate analysis code. You can thus record functional versions of your code and construct your analysis pipeline step by step:

https://git-scm.com

RStudio enables integration of your R or Python analysis codes into a logbook containing the analysis codes, the results obtained and a description of the methods used:

https://www.rstudio.com

Graphical representation of data is increasingly important: Beyond bar and line graphs: time for a new data presentation paradigm, Weissgerber et al., 2015.

A description of the main image problems observed in scientific publications in biology and recommendations on possible improvements in image processing is available: Creating clear and informative image-based figures for scientific publications, Jambor et al., 2021.

A video offering advice on how to improve the visualization of your data, to increase transparency and improve the reproducibility of your research work, is available from eLife, a non-profit organization that provides tools and documents to encourage more responsible behaviour.

Numerous tools exist to help people with no programming knowledge to create static and interactive graphics for scientific articles. These tools allow you to explore different ways of visualizing data and producing good quality graphs for publications.
These tools include:

Interactive dotplot lets you create dot plots, box plots, violin plots and combinations of these graphs,

Interactive line graph lets you create interactive line graphs,

Interactive repeated experiments dotplot lets you create interactive dotplots for independent repeated experiments.

Various publication methods are available. Full information is available on the CoopIST website, CIRAD, Delegation for Scientific Integrity.

For example, you can choose to publish a preprint of your manuscript before publishing in a peer-reviewed journal. Pre-publication servers are used to publish a preliminary version of your manuscript before submitting it to a peer-reviewed journal. Publishing in this way enables you to share your results and make your work more rapidly accessible.

A practical guide to preprints has been published by the Dutch Consortium of University Libraries and the National Library of the Netherlands (UKB), the Association of Universities of the Netherlands (VSNU) and the Netherlands Research Council (NWO). In this guide, you will find very practical information on the different preprint servers, how to publish a preprint, how to prepare a preprint, how to revise a preprint and how to link a preprint to an article published in a journal.

Preprint servers include:

BioRxiv, a preprint server for biology

MedRxiv, a preprint server for the health sciences

OSFPreprints from the Open Science Framework, for any scientific field

You can also choose to publish in an Open Access journal.

A list of Open Access journals that meet quality standards has been compiled by the Berlin Institute of Health. This list enables you to find Open Access journals relevant to your field of research, and provides information on average publication costs and time lapses for each journal.

The Directory of open access journals (DOAJ) offers a more in-depth and comprehensive database for open access journals in all fields of research. Journals have to be assessed for quality before being listed on DOAJ.

You can publish in an open archive even if you have not published in an open access journal (often after an embargo period). Useful information is available on the French Committee for Open Science’s website.
HAL is a multidisciplinary open archive for the deposit and dissemination of published and unpublished scientific articles and theses from French and international teaching and research establishments, and public and private laboratories. HAL-Inserm is the Inserm portal for this repository.

Take care to select high-quality journals and avoid “predatory journals”. Tens of thousands of journals exist, and new ones are launched every day. Among these, there are some 15,000 “predatory” journals: before submitting an article, it is essential to check the journal’s quality. A checklist for ensuring that the journal to which you plan to submit your article is a reputable one has been proposed by the international “Think check submit”. Similarly, Coop’IST offers tools to help detect predatory journals and publishers.

A video will give you a better understanding of the business side of scientific publishing.

Writing a manuscript also entails ensuring that it includes sufficient information for the research reported to be reproduced. Reporting guidelines can help. For example:

How to integrate reporting guidelines in your journal’s workflow

Ten simple rules for writing research papers, Zhang, 2014

The journal Nature provides assistance to answer your questions about sharing and making research data available.

A FAQ on research data policies is available on the Nature website: Research data policies FAQs.

There is also a practical guide for those who do not know where to deposit their data: Data repository guidance.

Finally, a set of resources on data availability statements can be found on the Nature website: Data availibility statement: guidance for authors and editors.

Inserm has published a short guide with recommendations on the use of the social media, designed for all staff members interested in expressing themselves and sharing their knowledge on the social media: “Guide pour un bon usage des médias sociaux“.

Almost every day, results of epidemiological surveys are brought to the attention of the public, commented on in the media or on the blogosphere, or used as a basis for political decisions. Inserm provides a practical guide for the “translation” of common and sometimes complex epidemiological concepts, which are defined and illustrated with concrete examples: Diffuser des connaissances en épidémiologie (Disseminating epidemiological knowledge).

You will also find resources on the Coop’IST and DoRANum websites.

Sharing research data is very useful because it enables data to be reused and individual data to be synthesised. This sharing is also important for improving the reproducibility of research. Many governments (including the French government), journals and scientific institutions encourage this sharing.

It should be noted that when it comes to data on individuals, specific rules apply, and a rigorous ethical analysis is required.

A few simple rules to make your data more useful:

An application, Shiny, has been designed by the QUEST centre (Charité-Berlin) to help biomedical scientists identify the most appropriate publication format for their data.

To deposit your open and citable research data in an appropriate archive, repositories allow you to download all the material related to your research project – so not only the publication but also the associated data or analysis scripts.

Zenodo – A general-purpose open-access repository developed as part of the European OpenAIRE programme and operated by CERN. This repository allows the submission of research documents, datasets, research software, reports and any other digital artefact related to a research project. For each submission, a persistent digital object identifier (DOI) is created, making the stored items easy to cite. There is no size limit.

https://zenodo.org/

Figshare – A versatile repository for non-traditional documents, data and search results. Files of up to 20 GB can be uploaded.

https://figshare.com/

Open Science Framework – A repository for saving documents and data from your research project before and after publication.

https://www.cos.io/products/osf

FAIR (Findability, Accessibility, Interoperability, and Reuse of digital assets) recommendations for the management of scientific data have been published and INSERM supports this movement. These recommendations aim to improve the retrievability, accessibility, interoperability and re-use of digital assets. The principles focus on machine actionability (i.e. the ability of IT systems to find, access, interact with and reuse data with no or minimal human intervention) because humans are increasingly relying on IT to process data due to the increasing volume, complexity and speed of data creation.

https://www.go-fair.org/fair-principles/

https://www.nature.com/articles/sdata201618

A guide, written by the University of Paris, is available online to help project managers draw up research: Data Management Plans (DMPs). The proposed structure takes into account the life cycle of data, from its creation as part of a research project to its archiving. It is also based on various data management plan models, such as that of the National Science Foundation (NSF) and the Interuniversity Consortium for Political and Social Research (ICPSR). In particular, this guide incorporates the expectations of the European Commission as part of Horizon 2020 and FAIR data management.

http://www.bu.parisdescartes.fr/doc/recherche/Realiser_un_DMP_V2_2018_Def_HRS4R.pdf