The traditional in vitro cell culture models use two dimensional (2D) techniques, employing isolated immortal cell lines originally obtained from people with cancer (25) and now banked and commercially available. The first cell line to be established was the HeLa cell line, which was obtained in 1951 from Henrietta Lacks, a 31yr old cervical cancer patient (26). 2D models are still largely used for many experiments that involve deciphering molecular mechanisms and drug discovery. However, there are limitations with these models, leading to partly accurate results and predictions. Although the 2D cell culture model does not resemble the actual tumor and overlooks the complex cellular architecture of cancer cells and tumor microenvironment, it is still the easiest and most economical way to perform large scale screening and pilot studies. In real world. cell division, maintenance of cellular homeostasis, and cell death all occur in a three-dimensional environment within the complex tissue architecture in higher organisms. Experiments with cells growing on a two-dimensional plastic surface cannot mimic the inherent intricacies of cell-cell interactions, cell-matrix interactions, and biomolecule availability and drug dose response (hormesis) (27). This led to the development of different techniques, trying to mimic the 3D tissue microenvironment, and these 3D approaches are being actively used to study diseases like diabetes (28, 29), cardiomyopathy (30), and cancer (31). Culture of cells in 3D, both with and without scaffolds, is revolutionizing biological research (32). The first research with cells in 3D dates to more than a century back when bacteria were clamped in a hanging drop (33). Later, this technique was employed to grow the first tissue in a scaffold-free manner (34). 3D cell culture models relevant to human disease started coming to the limelight since the 1980s and have been advancing since then. 3D cancer models resemble the tumor architecture more closely (35) and can mimic the cellular crosstalk in the different co-culture systems (36). Different cancer and non-cancer cell lines are grown in 3D scaffolds to resemble tumor microenvironment. Cytotoxicity testing, modelling of cancer stem cells, and epithelial to mesenchymal transition is better represented in the 3D culture system when compared to conventional 2D culture (31, 37). However, the major limiting factors of scaffold-dependent 3D cell culture models are its inability to mimic the biomolecular circulation and the controlled availability (which is physiologically aided by vasculature). Further, these scaffold-dependent models fail to mimic the appropriate biophysical extracellular structures/cues as necessary to tissues. To date, conventional 2D and 3D in vitro models fail to fully replicate the complex dynamic system operating within a tumor (38).

Other notable advancements in cancer research are the organoid and spheroid 3D cell culture systems. Both Spheroid and organoid models are 3D cell-culture models but with marked differences. A spheroid model is the simpler of the two, formed by of simple 3D aggregation of cells and does not contain any extracellular matrix or hydrogel scaffold and thus cultured as free floating aggregates. It can be developed from any cancerous or non-cancerous cell line or primary cells (by using a non-adherent plate). On the other hand, an organoid is a complex 3D structure of cells which is often supported by extracellular matrix or hydrogel scaffold and specific growth factors (32, 39) and is developed using stem cells primarily derived from the human healthy to tumor tissues. The cells in the spheroid predominantly has stem cell like property but, an organoid contains more differentiated cells (aided by scaffold and growth factors) and closely resembles the tissue architecture and function (40), which can be cryopreserved (41). Organoids grown with this method acquire a micro-anatomy, which is similar to the native tissue by self-organization and spatial orientation (42) and helps retain heterogeneity of the original tumor (19). Patient-derived organoid models have been generated with liver, colorectal, pancreatic, and prostate cancer (19). Gene-editing to introduce mutation and co-culture techniques to mimic different cellular interactions have also been recently possible using organoid models in cancer (43). The organoid models have advantages over the regular 3D model in simulating tumor-specific differential growth patterns like quiescence. On the other hand, the spheroid culture presents another 3D model, which resembles the 3D structure of a tumor and is established with low technical difficulty and cost. These spheroids loosely resemble the architecture of the native tumor and can also be co-cultured with other cells to mimic the exact tumor microenvironment (44). They have been shown to be more reliable for in vitro drug screening (45) and for modelling drug-resistance (46) than conventional 2D culture.

Features like hypoxic gradient and nutrition gradient can be mimicked more realistically in these advanced 3D models in comparison to the native tumor (47). Over the years, with the advancement in the knowledge of stem cells and the usage of patient-derived tumor organoid models, these advanced 3D models hold key promise in the field of personalized medicine (48–50). These characteristics make them more relevant compared to in vivo animal models, being not only cheaper but also lacking the ethical conflicts of the latter (51). Recently, there has been significant progress in the application of microfluidic-based devices to create models of human organs, specifically for drug development and bio-modelling. This is called the organ-on-a-chip/organ chip model, which accounts for the constant circulation of media, mimicking the human vascular system. Scientists have been successful in developing organ-chip models of intestines, blood-brain barrier, bone marrow, lungs, liver, kidney, and skin and using them as effective preclinical models in cancer, and other diseases (52). The organ chip model recapitulates tissue-tissue interface, and multicellular architecture very closely and helps modulate local molecular, chemical, and biophysical parameters in a precisely controlled manner (53). Tumor-chip models have been shown to closely resemble tumor stroma and are beneficial for modelling multi-organ metastasis (19). The organ-on-a-chip model has been used to study important cancer hallmarks like angiogenesis, migration, and invasion in a precise manner (53). Further, 3D based microfluidic research shows applications in the study involving drug testing, metastasis, and drug resistance (54). Although organ-chips are by far the most advanced of the in vitro technology systems, and successfully mimic organ-level physiology, this technology is still in its infancy and is likely to require more validation in different contexts with in vivo human system before we move to a paradigm shift in our methodologies for basic and translational cancer research.

Rodents are the main animal used as in vivo models to test drug efficacy. The xenograft mouse model is one of the most prominently used mouse models to model cancer and test drugs in which in most cases commercial human cancer cell lines are inserted within the immunocompromised mice (55). This strategy has been perceived as the gold standard for the preclinical drug testing and elucidating molecular mechanisms in cancer. However, these animal models have significant limitations. Since the model organisms are not genetically identical to humans, the tumors do not reflect the exact human condition and thus makes it difficult to predict drug efficacy and behavior. Furthermore, because the immune system is compromised, complex cellular interactions between the immune cells and the cancerous cells cannot be studied. Considering that we now know the role of immune cells in cancer cell survival and progression (56), these might not be the appropriate way to model human cancers (55). Furthermore, these cells which have been grown for many generations in plastic, may not be very ideal to recapitulate tumor close to being original. More recently, animal models have been used for patient-derived xenograft models (PDX) and genetically engineered models (GEM). In the PDX models, the primary tumor cells are taken directly from patients to develop a tumor in immunocompromised mice. It has added advantages of mimicking the heterogeneity of parental cancer. However, it also has the disadvantage of using immunocompromised animals (57) and that patient-derived cells have been shown to undergo animal-specific modifications (58), besides being expensive and time consuming. There have been several attempts to develop GEM to mimic the original human tumors (59). The mutations are either made in somatic or germline cells to model the tumor. Recent progress in gene editing tools has advanced this field considerably. These models are supposed to be superior to conventional models (59). However, the major problem of these models is that they do not incorporate human cells, and they are not cost-effective. An alternative to rodents, invertebrate (Drosophila) and fishes (zebrafish) have been used less prominently as a model for therapeutic research because of their lesser resemblance to the human genome. Naturally occurring tumors in canines and felines, are also used in cancer research. Although they have the advantage of naturally occurring tumors (60) and longer life span when compared to rodents, they have their own limitations. They do not exactly mimic human physiology. Additionally, because of the lack of scope of genetic modifications and the extra demand for time and efforts, lack of sufficient numbers, these models do not promise to be the most ideal for cancer research (61).

Besides the widely used traditional cell culture and animal models, human-specific models for cancer research offer exciting avenues for preclinical testing. Human tissues have been used in oncology research for a long time (62). The ongoing advancement in tissue and cell culture reveal their promise for overcoming the common problems of animal-based research. One promising aspect is the primary cell culture from patient samples, where tumors from the patients are directly used and cultured on a dish, maintaining the inherent cellular heterogeneity. This technique is more suitable for drug testing and modelling chemoresistance (63). In comparison to conventional scaffold-based 3D cell culture models, the spheroid and organoid models of the primary tumor are far better representative of the original human tumor and can be used for better cancer modelling. A more recent trend is to utilize the tumor cells in a co-culture system with other cells of the tumor stroma, such as fibroblast and immune cells. These models can be used for high throughput drug-screening assays and have been successfully tried in a 96-well culture system (45, 64, 65). Alongside, organ chip model has been successful in taking care of fluid circulation and biomechanics of the tissues very closely. These microfluidic devices perhaps provide the best platforms along with the advanced 3D models to build a better and more complete cancer model which can closely resemble the parental tumors (53, 66). The most advanced form of human tissue reconstruction is 3D bioprinting which is now applied to model different types of cancer. This field combines the principles of biomaterials and tissue engineering with the delicacy of cell biology to provide a printable tissue that can be used for organ transplantation, drug discovery, and personalized medicine (67). These models have shown promising results in mimicking native tissue microenvironment and complexities (68–70). 3D bio-printed cancer models showed greater cell survival, protein expression patterns that resemble host cancer, and higher chemoresistance to anticancer drugs similar to host. They can also retain features that closely resembled host tumor characteristics like tumor heterogeneity, necrotic cores, and microenvironment. These methods have helped to better understand cancer formation, progression, and response to anticancer therapies. Since animal models differ physiologically from human cancer, a bio-printed cancer model with human cells can be a better representative. These models are also able to mimic the effects of metastasis which is a significant advancement in the fields of in vitro cancer modelling. Bioprinted cancer models have been shown to provide effective reconstructions of host cell and cell-matrix interaction patterns. These models were not only able to simulate metastasis, but the local chemical signalling profile was also found to be similar to the host. The current limitation of this technique is that the cells can only be cultured for a limited period. The expansion of human cell biobanks will help in the supply of a variety of patient’s cancer cells (of different grades and types) to research institutes, which currently have limited accessibility. We discussed this further in the last section (Future roadmap and perspective).

Data generated from OMICs-based research approaches (such as genomics, transcriptomics, proteomics, and metabolomics) have added greater value to the field of cancer research as it has in any other biomedical field. The ‘OMICs’ generally refers to the analysis of structure, function, and origin of biomolecules derived from high throughput studies, along with their implication in the applied fields of biology (71). OMICs based studies tend to generate vast amounts of data, and a systematic approach is needed to store, access, and analyse them using various tools. Advances in the field of molecular and computational biology have enabled us to perform genomic, proteomic, and transcriptomic analyses from the patient samples or primary cell culture (72, 73). Databases generated from “Omics” studies certainly provide a platform for biologists to access them from different sources and integrate it in their research (74, 75). Although some of these databases are behind a paywall or require specific permission, most of these are in the public domain and are made freely available for cancer researchers. Some of the publicly available database document information regarding the human disease, mutations, expression of proteins, and response to drugs (e.g., TCGA Database) while other databases enlist similar information regarding cell lines (e.g., CCLE). Databases are also available to explore other potential areas like epigenetic modification, miRNA regulation, and detailed mutation analysis (76). Some of the relevant databases are listed in Table 1 , nevertheless describing them in detail is beyond the scope of this article. As we have discussed earlier regarding the therapeutic challenges posed by tumor heterogeneity, drug resistance, and drug mistargeting, it becomes vital to analyse OMICs data to understand disease progression, identify molecular markers, and response to therapy (98). The main use of cancer databases is to enable the generation of scientific hypotheses from the available datasets. These hypotheses can vary from the expression of a certain gene to the efficiency of certain drugs. In certain cases, these hypotheses can be tested using scientific tools, like cBioportal, which gives information about associated mutations, responsible protein function, and their association with survival. Alongside, proteomic databases are being used for developing diagnostics, identifying mechanisms, and designing treatment strategies. The main purposes of these databases are to save time by providing information that might be acquired from traditional wet lab and animal experiments, whereby avoiding repetition and duplication of experiments Moreover, many of these databases are based on integrated human and cell line experimental data and are carefully curated. Therefore, these databases can provide a starting platform where we can get information similar to that obtained with animal models and help to avoid further or duplicate use of animals. With data being generated at a high pace and their huge scope of application, one must admit that above-mentioned databases have been underutilized and can be used more broadly and efficiently.

The major advantage of many of the above mentioned repositories is that many of them contain patients’ data and thus are useful for accurate, human-relevant predictions. Increasingly, more data sets are being validated using other supporting experiments and have slowly started getting global acceptance. These approaches have the potential not only to complement but also to reduce many of the currently used wet-lab experimental models in an inexpensive manner. They open several new possibilities in the field of personalized medicine. However, a major limitation of these databases is their sheer volume and difficulty in handling them. The in silico models, which apply sophisticated algorithms, help in handling voluminous data of different cancer databases, to advance scientific understanding. Experience in computational analysis and specialized computational labs capable of handling big data are necessary to utilize the full potential of these databases. Therefore, a systematic approach is taken to analyse these data and model different outcomes. These can be built based on several approaches such as statistical models, network-based models, and tissue-based models (99). Examples of these models are provided in Table 2 .

A new in silico approach to drug development is based on AI (artificial intelligence). Machine learning and deep learning methods are employed to improve the above models (109). These models can automatically access most of the available data and draw a conclusion using the AI approach to design lead compounds, refine existing drugs, validate them on artificial disease models and predict the emergence of drug resistance (110). Some software include modelling options. For e.g., Cytoscape is a free software package that provides user-friendly visualization of these open-source data (111). It may be noted that several of the primary databases also integrate these modelling features or allow the users to model the available data according to their needs. Though we have described these platforms primarily as cancer databases, it must be noted that several of these databases serve a dual role of data repository as well as a platform to analyse these data and model cancer. The field is evolving rapidly, and very soon AI-based platforms will be available as a complete package. These packages may be able to offer a more user friendly approach to analysing the databases and thus, supplementing cancer research. Advanced AI platforms, in combination with superior 3D cancer models, have the potential to replace unnecessary animal experiments as they more human relevant besides being faster, resource saving, effective and reproducible.