The objectives of the development of GIATE are as follows:

• To provide a consistent information framework for reporting therapy experiments in a transparent way

• To support efficient data-sharing of therapy experiments' methods and results

• To guide the description of therapy experiments, enabling reproducibility, data reuse and re-purposing to avoid duplicated effort, support comparisons between experiments and increase the quality of the data

• To facilitate the integration of data coming from more than one experiment in a machine-processable way

• To allow for the possibility of obtaining aggregated and new knowledge coming from the statistical analysis of the data or mining of the aggregated information (data and knowledge mining)

• To develop tools to help the different GIATE stakeholders to use the GIATE guidelines

• To facilitate semantic publication of therapy experiments 30

The scope of the GIATE guidelines is therapy experiments, such as small molecules targeting molecular pathways, engineered protein therapeutics, radio-immunotherapy, anti-vascular therapies, cellular therapies and others. For the design and development of new and improved therapies, it is crucial to be able to integrate laboratory and clinical data. Thus, GIATE aims at providing an information framework to record the properties of the target, the agent and models, the therapy investigation design, context and outcomes from the different studies.

These guidelines are mainly intended for use by researchers designing a therapy experiment, to assist them in collecting consistent data elements. Moreover, they are intended to act as a reference in order to form a coherent description of an experiment.

GIATE is also intended to be used by journal editors and referees to help confirm that certain data elements are uniformly presented in publications, and that their interpretation is unambiguous, allowing for the verification of the conclusions obtained.

Other potential beneficiaries of GIATE are the communities that regulate and fund therapy experiments. Within these communities, the Antibody Society 22 has adopted the GIATE specification.

Additionally, GIATE is intended to be used for educational or training purposes: we endorse the view that good research reporting habits should be introduced as early as possible in a researcher's career, to both students and young researchers 9 . The structure of GIATE presents the main properties of a therapy in a consistent manner, which facilitates the understanding of the therapy's strategy, intentions and achieved outcomes.

To benefit the different stakeholders, the GIATE project aims at providing tools that will help in making the extra effort of recording the experiments minimal and worthwhile.

GIATE 31 is part of the MIBBI (Minimum Information for Biological and Biomedical Investigations) Consortium 6 32 . Following the MIBBI design guidelines used in other specifications 21 33 , GIATE does not intend to provide an exhaustive list of data required or resulting from a therapy experiment. Instead, GIATE balances the trade-off between the depth of information required to record an experiment and the burden for the researcher to produce and maintain this metadata (or data about the data). Thus, the compromise between sufficiency and practicability 33 has been considered in GIATE's development.

GIATE is structured as a set of modules, most of which are specific to therapy experiments. Others, such as the citation module, are generic and could be reused within other guidelines. GIATE establishes the relationship between its constituent modules. Each module encapsulates the information related to a specific sub-domain of therapy development. A modular design guarantees that changes in a specific module will be local and will not affect the guidelines and tools related to other modules.

In comparison to guidelines that focus on specific assay results, such as MIAME 18 34 or the Minimum Information About Cellular Assay (MIACA) 35 , GIATE can be considered a 'meta-guideline' because it deals with high-level information from studies ranging from the lab to the clinic. These other guidelines form independent modules in GIATE, where appropriate. Users are therefore presented with domain specific guidelines if a higher level of detail is required to record any aspect of their data. GIATE, then, is the 'glue' linking diverse data to support translational research.

Another important feature of GIATE is that it is not static. As with other checklists such as the Minimum Information for Molecular Interaction experiments (MIMIx) 21 , it is expected that through the increasing participation of the therapy community, GIATE will evolve to reflect the changing requirements in the context of a rapidly evolving therapeutics science. In this paper we present specific versions of the GIATE checklist and its related tools.

Figure 1 depicts a schematic view of the main components of a therapy experiment, which are classified into the GIATE core modules, supplementary modules and links to external resources.

As per MIBBI guidelines, we consider the distinction between Investigation, Study, and Assay. An investigation refers to 'a self-contained unit of scientific enquiry' 6 that is characterised by an hypothesis or objective and a design, which is defined by the relationship between one or more studies and assays. Core GIATE modules can be seen as a tree, including the therapeutic investigation description at the root and more specific studies including data on the therapy development at the branches and leaves.

The main GIATE module is the Therapeutic Investigation, whose design is determined by the Therapeutic Target and the Therapeutic Agent. In turn, the Agent may be composed of one or more Components.

Figure 2 presents the internal structure of the Therapeutic Investigation module, with sub-modules describing the target and the agent with its components. Supplementary modules for citations about the target, agent and components are also included, together with links to external resources.

The studies are represented by the different Models (see Figure 1) as each investigation may be applied to one or more models. Model types include: Molecular, Cellular, Animal (or pre-clinical) and Clinical. A particular Investigation may have been applied only to some of the models, for example to Cellular and Molecular models but not to the rest. Progress made from bench-side to bed-side can be tracked with the information in GIATE element types. Common characteristics of all the models are grouped into a generic Model module. Each of these models may have one or more assays. For example, a cellular model might contain information about cellular assays, which are reported to the MIACA checklist. When describing each of the modules, we discuss some of the relevant guidelines that researchers should consider for each sub-domain. Figure 3 shows some of the relevant guidelines per each module.

In addition to the GIATE core modules, we designed a module representing Citations, which is described in more detail when the GIATE checklist is introduced. In the future, other supplementary modules such as Imaging will be considered, given their role in therapy development 36 .

As described in the background section, GIATE was initially designed as a set of Common Data Elements (CDEs), as per the ISO/IEC 11179 standard for metadata registries 37 . A registry not only specifies the content that it maintains, but also the rules, operations and procedures that it uses to maintain its content. According to the registry standard, the set of CDEs determines a classification scheme, as they are grouped by the common characteristic of representing GIATE information.

Figure 4 shows a schematic view of the components of the ISO/IEC 11179 standard 37 . A data element is the basic container for data and it might represent an abstraction or an entity from some system. Data elements have both representational and semantic components 37 . In turn, the semantics involve two aspects: symbolic and contextual types. The contextual semantics comprise a data element concept, which indicates the types and characteristics of objects for which data are recorded 37 . The symbolic semantics come from a conceptual domain, which is a set of categories (enumerated or expressed with a description) representing the permissible or allowable values in a value domain. The representational level includes the data element itself as well as one or more associated value domains, specifying the set of permitted values 37 . We note that the metadata registry standard includes in a single model conceptual and representational aspects. The content to report is determined by a data element as an ObjectClass, a Property and a Value Domain.

Thus, a CDE involves simultaneously the three aspects of a reporting structure as seen before: what to record, how to record it, what is the meaning of the information recorded. Additionally, what to record (checklist) and how to do it (format) are intertwined between the conceptual and representational levels.

When identifying the sets of CDEs for a particular domain, it is recommended to reuse existing CDEs as much as possible, as this results in increasing interoperability of new data resources based on the new classification scheme with existing data resources. When developing the GIATE checklist we found that making this effort to reuse CDEs could impose constraints on the content of the key elements. For example, when dealing with the animal model some of the existing CDEs had an ObjectClass related to Animal while others had an ObjectClass related to Organism: Animal Cancer Model Phenotype Description java.lang.String and Organism Species Name java.lang.String 38 . However, when specifying what to record, having to use these two CDEs might be confusing, as in both cases we are referring to properties of the Organism used in the Animal Model.

In this paper, then, we present the key information elements independently of the CDEs, which can be associated a later stage. Thus, we divide GIATE in the three levels as determined by a reporting structure, and present the content to be reported independently of any data format.

One of the tools developed to support GIATE is the GIATE Notebook-a piece of software that can be used as an electronic lab book to capture data on therapy experiments. The interface is composed of three panels: one for the GIATE elements, another one containing the CDE details showing associated terminology for each element and the third for data entry. The data produced with the GIATE Notebook can be exported as an eXtensible Markup Language (XML) document or in Portable Document Format (PDF).

More details on the GIATE notebook and its use for the ADEPT therapeutic investigation applied to an animal model 39 were presented in 28 .

We have developed a checklist with the key information that should be recorded about therapy experiments. The checklist main modules are as in Figure 1. In this paper, we will briefly describe each of the modules and will show the module concerning the clinical model in more detail, as this is the main component of the use case presented at section 2.3. The complete GIATE checklist file, version 0.1, is available as Additional file 1.

In the therapeutic investigation module (see Figure 6), we have included general information about the investigation: its goals, description, therapy type, experimental factors and its conclusions.

Additionally, the target (CD25, Interleukin-2 receptor subunit alpha), the agent (¹³¹Iodine labeled CHT-25 chimeric antibody) and its two components (CHT-25 and ¹³¹Iodine) are specified. CHT25 is a chimeric monoclonal antibody with murine variable regions and human constant regions. CHT25 was radiolabelled with ¹³¹I, an appropriate radionuclide for radio-immunotherapy as it has β emission length of 0.8 mm and γ emissions for imaging purposes.

The citation module is used to link to papers describing properties of the target 63 64 and the method used for the radioiodination of the antibody in the agent 65 , as referred in the original paper.

The affinity of binding between CHT25 and IL-2 receptor is approximately that of IL-2 itself 62 . Previous results refer to the unlabeled antibody which has been used to prevent transplant rejection in renal patients. An alternative unlabeled antibody has shown short-term benefit in human T-cell lymphotrophic virus-associated lymphoma, where the IL-2 forms part of a growth pathway 62 .

No cell lines studies exist for this therapy.

There is no suitable representative animal model for ¹³¹I-CHT25. While Rhesus monkeys contain the same IL-2R epitope, they are not suitable for therapy studies. Toxicology for ¹³¹I-CHT25 has not been performed in pre-clinical models either 62 .

Thus, within the reporting for the CHT-25 study the animal model includes comments about these facts.

The first section of the clinical model in GIATE-TAB includes general information about it:

• The objectives and endpoints of the CHT-25 study such as evaluation of the toxicity, pharmacokinetics, immunogenicity and anti-tumour activity of CHT25

• The number of patients, with description of the eligibility and exclusion criteria: the study involved 14 patients who had CD25 positive lymphomas (Hodgkin lymphoma, HTLV associated adult T-cell lymphoma and peripheral T-cell lymphoma) in which standard therapies had failed or were not tolerated 62 .

• The study design or type, which in this case is single centre, open label, non-randomized, multiple dose escalation phase I trial.

• The conclusions: it was found that CHT25 has an important clinical activity in CD25 positive refractory lymphomas; it is relatively non-immunogenic with low toxicity at a non-myeloablative dose. Further studies are required to assess clinical effectiveness and these will be carried out in a Phase 2 trial.

This sub-module on General Information is linked to a citation module referring to the article 62 .

For this therapy, information was included about individual patients. Elements referring to the type of lymphoma and treatment history are considered, including treatments such as chemotherapy, Autologous Stem Cell Transplant (ASCT), radiotherapy, time since last therapy, stage at therapy, and bone marrow involvement.

The sub-modules for genetic/epigenetic profile, target distribution and pharmacodynamics are not relevant for this particular investigation, and that is indicated in GIATE-TAB.

Information for the dose regimen is included for the investigation and for individual patients. The study consisted of a dose escalation using a standard dose of 10 mg of CHT25 antibody, with escalation of the radioactive iodine from 370Mbq/m ² to 2960Mbq/m ².

CHT-25 was administered to 13 patients in 24 cycles. The dose limiting toxicity was determined at 2960 Mbq/m² with grade 4 myelosuppression in one patient. The patient failed stem cell re-engraftment and died of infection. The dose was reduced to obtain the maximum tolerated dose and 3 patients were treated at 1200 Mbq/m² with recruitment completing at 1480 Mbq/m². Other toxicities were mild.

A distribution study was performed to analyse radioactivity uptake in target and non-target tissues. The main conclusions are summarised in GIATE-TAB.

Details of the radiation dosimetry study are also included. It is noted that this GIATE sub-module is only relevant for radioimmunotherapy experiments.

GIATE-TAB also includes information about Pharmacokinetics (PK) studies, i.e. how a drug or substance is absorbed, distributed, processed and eliminated in animals and humans. In order to study the PK for ¹³¹ I in the CHT-25 investigation, blood samples were taken into EDTA blood tubes at the following time points, when possible: 1, 3, 6 and 24 h, then on day 2, 3, 6 and 9. The data presented in the paper 62 has been transcribed to GIATE-TAB including general parameters (e.g. median clearance for 50% and 90%) as well as per patient information. The latter comes from a tabular representation in the paper, giving the parameters of the PK interpolation curves per patient. The curve is either monoexponential, described by a single parameter, or biexponential, described by two parameters.

Finally, a sub-module indicating the outcome is also included, at the investigation and patient levels. Elements included are best response, Common Toxicity Adverse Events Grade and survival status. As the Cheson criteria 66 were used to classify the patient response (best response), e.g. as stable disease (SD), complete response (CR), partial response (PR) and so on. A citation module referring to Cheson et al. 's article 66 was linked to the outcome study sub-module.

In this section, we indicate how GIATE has contributed to knowledge about CHT-25 and how having the data elements in the GIATE-TAB spreadsheet will help in understanding the different components of the therapy as well as facilitate secondary use of the data.

Firstly, the spreadsheet provides an overall view of the CHT-25 therapeutic investigation, which highlights the main points and their relationships. This process is simplified by the spreadsheet in comparison with the more time consuming task of reading the paper, the protocol, and if necessary, contacting the authors of the trial.

Secondly, the spreadsheet provides links to external resources that are available neither in the scientific article nor in the protocol. For example, the GIATE-TAB format for CHT-25 makes clear that the therapeutic target is CD25-Interleukin-2 receptor subunit alpha, accessible in UniProt (at http://www.uniprot.org/uniprot/P01589, version number 130). The specific link to UniProt allows users to uniquely identify the molecule that is mentioned in the paper. Thus, scientists wanting to analyse the CHT-25 trial could navigate to extra information about the CD25 molecule. Similarly, additional information about the agent component CHT-25 is accessible through the DrugBank database (at http://www.drugbank.ca/drugs/DB00074).

Thirdly, the paper provides static information presented in diagrams and tables. On the other hand, the GIATE-TAB for CHT-25 makes re-use of the information in a dynamic fashion. For example, while the pharmacokinetics analysis is available in the paper as Table 1 62 , the figures cannot be immediately used to generate the interpolation curves. By having the data in GIATE-TAB, it is possible to generate these curves dynamically for further comparison and analysis of the pharmacokinetics data.

Finally, having the CHT-25 data in GIATE-TAB format facilitates answering queries about the therapeutic investigation much easier than having to go through the whole paper or protocol. For example, GIATE-TAB allows users to identify quickly what were the studies performed for the CHT-25 therapy and compare the Cheson score for the outcome of each patient at a glance. Additionally, GIATE-TAB is a step towards answering queries about therapeutic investigations in a machine processable way. As part of our future work, we intend to build a therapeutic investigations knowledge base, which will support to retrieve this kind of information.