• Artificial intelligence is used in toxicology to predict the effect of a chemical on a living organism. The main applications are machine learning and deep learning. (tile n6)
• Attend a congress. Even without a poster or a lecture to give, a congress is a perfect way to share with experts and cultivate your network.

• Case studies in toxicology are in-depth analysis on a specific subset of chemicals or exposure situations which can prove or strengthen a scientific conclusion. For example, suppose a theory claims that chemicals in water lead to neurological defects in fish. In that case, a case study on a specific water region where such chemicals are present in known concentrations could support that theory and explain why. Case studies are particularly useful to support the validation of a method. (tile n39) validation of a method.
• A 2D cell system is the classical method of cultivating cells on a flat surface. Cells from many human/animal organs can be grown and used to test chemical/substance toxicity. (tile n8)
• A 3D cell system is a modern technique where cells can grow in all directions thanks to the presence of specific scaffolds in the dish. In such a setting, cells create structures similar to real organs, which can be used to test chemical/substance toxicity. (tile n9)
• Collaboration with national regulatory agencies: When developing a new method, it is beneficial to include the regulatory agencies in the process since they will be the ones deciding whether the technique can be used. (tile n44)

• Data analysis is the process of systematically applying statistical and/or logical techniques to describe,illustrate, condense,recap and evaluate data. (tile n29)
• Data curation and cleaning is the process of collecting all the information related to these data, such as how they have been collected, where they have been stored and how others can reuse them. (tile n16)
• A data management plan is a series of procedures to collect and store scientific data in a standardised way. Detailed information on how the data have been generated and catalogued help other scientists exploit them for new studies which improves scientific reproducibility. (tile n12)
• A defined unambiguous algorithm refers to an algorithm that is both clear and straightforward to the point that any person can follow its steps and reach the correct conclusions. (tile n30)
• A defined endpoint in toxicology is the result of a study conducted to determine how dangerous a substance is. The data collected are often used to report the substance’s toxicity to regulatory agencies. Endpoints can include mortality/lethality, behaviour, reproductive status or physiological and biochemical changes. (tile n14)
• Define acceptance criteria is the process of specifying which characteristics/parameters/results a method must present to be accepted and used. For example, in toxicology, it could mean being able to detect a substance at a specific concentration or to predict a number of effects on the human body for a chemical. (tile n26)
• Defined domain of applicability refers to an essential step in mathematical models such as QSAR during which the chemicals and situations that the model can work on are specified. It is based on the information used to create the model itself; for example, if a QSAR is developed from data on chemicals found in water, it cannot be used to predict chemicals found in the air. (tile n35)
• Defined mechanistic interpretation refers to a direct causal link between a chemical and an observed toxic effect on organisms or the environment. This is of great importance when considering regulatory toxicology. (tile n36)
• Deep learning is a machine learning method which uses artificial neural networks to mimic human brain processes. In toxicology, it is used to predict the effect of substances on living organisms. (tile n3)

• Endpoint-specific control (EC) shows that pathways considered to be relevant for test endpoints are indeed affecting the test endpoint. EC helps correlate (by concentration and time) compound effect on pathway (activity measure to be established) and on test endpoint (standard test readout). (tile n23)

• Endorsement of an AOP (Adverse Outcome Pathway). An AOP describes the various steps, from exposure to a particular chemical, leading to a final toxic adverse effect on an organism or the environment. Each step is usually a molecular event (chemical binding a receptor or damage to the DNA), which is necessarily triggered in order to move to the next step, like the domino effect. (tile n40)

• EU Parliament meeting. Elected policy makers have the power to co-design policies that affect how chemicals are regulated and assessed. These decisions can be based on scientific evidence, and it is crucial to learn how to discuss with policy makers to understand what they need and expect. (tile n46)
• EU Agency meeting: Several agencies in the European Union work on the implementation of the chemical regulations (e.g: ECHA, EFSA, EMA). Discussing with their experts and desk officers can helpdesigning a better method. (tile n47)
• An external validation consists of assessing whether a test method developed in the historical laboratory will generate the same results for the same substances in another laboratory with a different team of scientists. (tile n19).

• Guidance documents are prepared by regulatory agencies to illustrate how an endpoint should be assessed. For example, how to report information on a new chemical for industries to national and international agencies. (tile n41)
• Guidance document EFSA – ECHA: see “Guidance documents”above. (tile n42)

• In vitro in vivo extrapolation (IVIVE) means translating an experimental result obtained in vitro into an in vivo setting. An example is to measure how fast a chemical is eliminated by human liver cells to estimate how long it will stay in the human body. (tile n27)
• In vitro & in vivo data can be extracted from existing data set and being used to build in silico models or during a validation study.

In vitro methods are laboratory techniques based on the use of cells and other biological materials which do not involve the use of entire living organisms such as animals.

In silico methods in toxicology are technologies based on software, algorithms and mathematical models used to predict the effect of chemicals on living organisms.

• A lecture in a conference is where scientific research is presented in front of peers (scientists and experts). Communicating on your method development is a great way to challenge it and reach better results or spring collaborations. (tile n49) or spring collaborations.

• Machine learning is used in toxicology to predict how a substance will affect various organisms, such as humans and animals based on the physical and chemical characteristics of the substance in question combined with the results of in vitro and in silico methods. (tile n2)
• Medium FBS-free is a cell culture reagent which does not contain animal-derived foetal bovine serum (FBS). FBS is still an important component of the laboratory solutions used to grow cells. However, it is extracted from bovine foetuses and its composition varies from batch to batch, introducing potential reproducibility issues. (tile n11)
• A workflow model (modelling workflow) is the sequential series of steps that make up a process. For example, in toxicology, it can describe how various computational and statistical methods, such as QSAR or PBPK, work. (tile n17)

• Model verification and validation starts after functional specifications have been documented and initial model development has been completed. Model verification and validation is an iterative process that takes place throughout the development of a model.(tile n31)