Modeling tumor evolution

Abstract

The present invention relates in particular to a computer-implemented method for simulating the evolution of a tumor associated with an oncogene. The method comprises providing (S10) a plurality of data segments, each data segment corresponding to a given cell of the tumor and comprising the degree of activation of the oncogene in the given cell. The method further comprises providing (S20) a model configured to take an input data segment and output information about the proliferation of the corresponding given cell corresponding to the input data segment. The information about the proliferation depends on the degree of activation of the oncogene. The method further comprises running (S30) the model on one or more data segments of the plurality of data segments and updating (S40) the plurality of data segments based on the results of running (S30). This method improves the simulation of the evolution of the tumor.

Description

Technical Field

[0001] The present invention relates to the field of computer programs and systems, and more particularly to methods, systems and programs for simulating the evolution of tumors. Background Art

[0002] There are computer-implemented methods for simulating the evolution of tumors. With the increase of computing resources, these methods allow the dynamics of tumors to be studied and thus form an important tool and source of information in the medical field.

[0003] Some of these methods are based on solving partial differential equations. These methods allow simulating tumors at specific locations and describing tumor behavior at the tissue scale. However, although these models can describe tumors, their accuracy is relatively low.

[0004] In this context, there remains a need for improved methods for modeling the evolution of tumors. Summary of the invention

[0005] Therefore, a computer-implemented method for simulating the evolution of a tumor is provided. The tumor is associated with an oncogene. The method includes providing a plurality of data segments. Each data segment corresponds to a given cell of the tumor. Each data segment includes the degree of activation of the oncogene in the given cell corresponding to the corresponding data segment. The method also includes providing a model. The model is configured to obtain an input data segment. The model is also configured to output information about the proliferation of the corresponding given cell corresponding to the input data segment. The information about the proliferation depends on the degree of activation of the oncogene in the corresponding given cell. The method also includes running the model on one or more of the plurality of data segments. The model also includes updating the plurality of data segments based on the results of the running.

[0006] The method may include one or more of the following:

[0007] - Each data segment also includes the spatial localization of a given cell of said tumor;

[0008] - the information about the proliferation of the corresponding given cell comprises information about the presence or absence of division of the corresponding given cell, and when the information about the proliferation of the corresponding given cell comprises information about the presence of division, the information about the proliferation of the corresponding given cell further comprises information about the corresponding spatial location for each cell generated by the division, each corresponding spatial location being near the spatial location of the corresponding given cell;

[0009] - said information about the proliferation of said respective given cell also depends on the availability of space in the vicinity of said respective given cell;

[0010] - the information about the presence or absence of division of the respective given cell is based on the spatial availability in the vicinity of the respective given cell, and when the information about the proliferation of the respective given cell comprises information about the presence of division, the respective spatial location in the vicinity of the spatial location of the respective given cell is an unoccupied spatial location;

[0011] - when the information about the proliferation of the corresponding given cell includes information about the presence of division, the information about the proliferation of the corresponding given cell also includes the corresponding degree of activation of the oncogene of each cell generated by the division based on the degree of activation of the oncogene of the corresponding given cell;

[0012] - a change in the respective degree of activation of said oncogene of each cell resulting from said division corresponding to at least a partial probability of the degree of activation of said oncogene of said respective given cell;

[0013] - the model is decision-based and includes, with respect to the respective given cell, a decision about death or survival, a decision about the presence or absence of division, and a decision about the acquisition of one or more new changes during division;

[0014] - providing the type of treatment on which the information on proliferation depends;

[0015] - iterating the providing, running and updating (S50), and the plurality of data segments provided in each iteration are the updated plurality of data segments of the previous iteration;

[0016] - providing data relating to a patient suffering from said tumour, said model comprising parameters depending on the data relating to said patient;

[0017] - providing the model comprises: determining the parameter to be configured to retrieve the values ​​of the plurality of data segments by: defining one or more other data segments, each other data segment corresponding to a given cell in the initial state of the tumor, each other data segment comprising the activation degree of an oncogene in a given cell; running the model on the one or more other data segments; and updating the plurality of other data segments based on the results of the running;

[0018] - optionally, initializing the value of a parameter based on data relating to said patient; and / or

[0019] - Determining statistical data related to the oncogene based on the updated plurality of data segments.

[0020] A computer program is also provided, comprising instructions for performing the method.

[0021] A data structure comprising the model is also provided. In other words, the data structure comprises a specification of a model, the model being configured to obtain an input data segment corresponding to a given cell of a tumor and comprising a degree of activation of an oncogene associated with the tumor in the given cell. The model is further configured to output information about the proliferation of the given cell, the information about the proliferation being dependent on the degree of activation of the oncogene in the given cell. The model is configured to be run.

[0022] A computer-readable storage medium is also provided, on which a computer program and / or a data structure including the model is recorded.

[0023] A system is also provided that includes a processor and a graphical user interface coupled to a memory having recorded thereon a computer program and / or a data structure including the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Embodiments of the invention will now be described by way of non-limiting examples and with reference to the accompanying drawings, in which:

[0025] Figure 1 A flowchart illustrating an example of a method;

[0026] Figure 2 A flow chart showing an example of adjusting parameters of a model using patient-related data;

[0027] Figure 3 A diagram showing an example of a method including a calibration model;

[0028] Figure 4 An example of an image of a tumor cell being analyzed for constructing a plurality of data segments is shown;

[0029] Figure 5A and 5B a flow chart showing an example of a model of the method;

[0030] Figure 6 An example of a GUI for entering patient-related data is shown;

[0031] Figure 7 An example of a simulated tumor is shown;

[0032] Figure 8-9 Examples of statistics calculated from the model are shown;

[0033] Fig.10 An example of a system is shown;

[0034] Fig.11 a flow chart showing the data flow of an example of a model;

[0035] Fig.12 A flow chart showing the data flow of an example of a model including several iterations;

[0036] Fig.13 A flow chart showing an example of determining inputs and parameters of a model; and

[0037] Fig.14 A flow chart showing an example of calculating personalized parameters of a model for a patient. DETAILED DESCRIPTION

[0038] refer to Figure 1 A flowchart is provided, which proposes a computer-implemented method for simulating the evolution of tumors associated with oncogenes. The method includes S10: providing multiple data segments. Each data segment is information stored in a computer, which corresponds to a given cell of a tumor. Each data segment includes the degree of activation of an oncogene in a given cell. The method also includes S20: providing a model configured to take the data segment as input. The model is also configured to output information about the proliferation of the corresponding given cell corresponding to the input data segment. The information about proliferation depends on the degree of activation of the oncogene in the corresponding given cell. The method also includes S30: running the model on one or more data segments, which are part of multiple data segments. The method also includes S40: updating multiple data segments based on the results of the model running.

[0039] This approach improves the simulation of tumor evolution.

[0040] Notably, the method simulates the evolution of a tumor by calculating the proliferation of one or more cells at S30 and updating the plurality of data segments provided at S10 based on the result of S30 at S40 .

[0041] Since a model configured to output proliferation information according to the degree of activation of oncogenes is provided at S20, the model takes into account the degree of activation of oncogenes when simulating how cells proliferate. This improves the accuracy of the simulation of the evolution of tumors. In fact, how cells proliferate is affected by the degree of activation of oncogenes in cells. For example, as the degree of activation of oncogenes in cells increases, the division rate of such cells will increase.

[0042] Since the proliferation information of the cell corresponding to the data segment used as the model input is outputted depending on the degree of activation of the oncogene in the cell, the simulation of the evolution of the tumor takes into account the information at the cellular level. This improves the accuracy.

[0043] Since each data segment from the plurality of data segments includes the activation degree of the oncogene in the corresponding cell, and the updating S40 of the plurality of data segments is based on the result of running S30, each of the updated plurality of data segments includes the activation degree of the oncogene in its corresponding cell. Therefore, the intra-tumor heterogeneity related to the activation degree of the oncogene in different cells is maintained throughout the simulation process. This also improves the accuracy of the simulation. In fact, cancer is a heterogeneous disease, and knowledge of intra-tumor heterogeneity improves the accuracy of diagnosis and treatment, such as breast cancer.

[0044] The evolution of a tumor is how the tumor develops over time. For example, the size of a tumor may increase over time and as tumor cells divide. Alternatively, if a patient with a tumor receives treatment, the size of the tumor may decrease. Additionally or alternatively, the evolution of a tumor may include evolution of tumor cells, such as acquisition of resistance of one or more cells to certain treatments and / or changes in the activation level of one or more oncogenes in one or more cells.

[0045] Tumors are associated with oncogenes. Therefore, oncogenes may be active in at least one cell of a tumor. Oncogenes in such cells may have varying degrees of activation. In an example, the degree of activation of an oncogene may correspond to the number of oncogenes in a cell. Additionally or alternatively, the degree of activation of an oncogene may correspond to the expression level of an oncogene in a cell. The degree of activation of an oncogene in a cell may be the result of one or more changes in the cell. In an example, the change may include any one or any combination of: an increase in gene copy number, an activating mutation, and / or an interchromosomal rearrangement or intrachromosomal rearrangement resulting in a gene fusion.

[0046] The multiple data segments provided at S10 correspond to given (i.e., corresponding / corresponding) tumor cells, respectively. One data segment corresponds to one tumor cell. Each data segment includes the degree of activation of an oncogene in its corresponding cell. In other words, each data segment includes a value representing the actual degree of activation of an oncogene in a corresponding cell, and may represent this real degree of activation approximately and / or by simulation. In an example, multiple data segments may each include the degree of activation of more than one oncogene associated with a tumor. The degree of activation of more than one oncogene in one data segment may come from different types of oncogenes. In an example, two data segments may include the degree of activation of different oncogenes.

[0047] The tumor cells corresponding to each of the multiple data segments represent real tumor cells. In an example, the multiple data segments provided at S10 can be determined using available information about a person's tumor cells. The person can be a patient suffering from cancer. In an example, each data segment can be determined using available information about a corresponding single tumor cell. In this case, the degree of activation of oncogenes in cells included in the multiple data segments provided at S20 can correspond to the actual degree of activation of oncogenes in cells belonging to the patient's tumor at a given point in time. Alternatively, the degree of activation of oncogenes in one or more data segments can be inferred based on the available information about tumor cells. In an example, the distribution of the degree of activation of oncogenes in the data segments of the multiple data segments can correspond to the actual distribution of the degree of activation of oncogenes in a cancer cell population.

[0048] The proliferation information output by the model provided at S20 may be any information representing the temporal evolution of the cell (i.e., how the cell proliferates) and / or the temporal evolution of the mechanism cell that affects the degree of activation of the oncogene in the cell. In an example, the information about proliferation may include information about cell death or survival (i.e., information about whether the cell survives), information about the absence or presence of division (i.e., information about whether the cell divides and when it survives), and / or information about the evolution of the degree of oncogene activation (i.e., information about whether the degree of oncogene activation in two cells generated by the division of a parent cell has changed relative to the parent cell, such as, for example, information about the acquisition of one or more new changes during division).

[0049] The information about cell proliferation output by the model depends on the degree of activation of oncogenes in the cell. The dependency can be any relationship that implements knowledge about real cell processes. For example, it is known that if the degree of activation of oncogenes in a cell increases, the division rate of the cell corresponding to the data segment as input will increase. In the example, the dependency can be such that as the degree of activation of oncogenes in the corresponding given cells input to the model increases, the corresponding given cells are more likely to divide and / or more likely to increase the degree of activation of oncogenes in the two cells produced by the division relative to the degree of activation of oncogenes in the corresponding given cells (e.g., more likely to obtain one or more new changes during division). Therefore, the model provides accurate output.

[0050] The model is run on one or more data segments of the plurality of data segments S30. The model simulates the evolution of one or more cells, each cell having a corresponding degree of activation of an oncogene associated with a tumor. The collective evolution of the cells simulates the evolution of the tumor. In an example, the simulation may be initiated from a single cell. Alternatively, the simulation may be initiated from a group of cells.

[0051] The multiple data segments updated at S40 represent the evolution of the tumor. The updated multiple data segments represent cells corresponding to the multiple data segments provided at S10 after running in S30 at a later time point. In this way, the corresponding activation levels of the oncogenes included in the updated multiple data segments correspond to the activation levels of the oncogenes in the corresponding cells at the later time point. If the activation level of the oncogenes in the cells represented by the at least one data segment does not change after simulating the evolution of the cells, at least one data segment may remain unchanged after the update. Alternatively or additionally, at least one data segment may be modified after the simulation. In an example, if after running S30, the information about the proliferation of the cells corresponding to the data segment includes information related to cell death, at least one data segment may be discarded. During the update S40, the discarded data segments may be deleted from the multiple data segments. Alternatively or additionally, a new data segment may be created for a cell that is generated by the division of the cell corresponding to the data segment as input during running S30. In this case, a value may be set. Multiple data segments may be updated after each run of the model. Alternatively, the model may be run on the plurality of data segments before updating the plurality of data segments with each corresponding information regarding cell proliferation.

[0052] The change between the provided plurality of data segments and the updated plurality of data segments simulates the evolution of the tumor. Based on the simulation of cell proliferation performed during operation S30, each updated data segment has an updated activation level of an oncogene associated with the tumor. The evolution of the simulated tumor maintains the simulated heterogeneity reflected by the different activation levels of oncogenes in different cells. Therefore, the method allows the prediction of the evolution of heterogeneity between tumor cells to be determined.

[0053] In the example, each data segment may further include a spatial location of a cell corresponding to the corresponding data segment. Spatial location allows cells to be located in a two-dimensional or three-dimensional space. The two-dimensional or three-dimensional space may have a maximum size. In other words, according to the maximum length of a direction in the three-dimensional or two-dimensional space. In the example, the spatial location may be a set of coordinates corresponding to a point, a surface, or a volume. In the example, the spatial location may be specified with respect to a grid composed of cells (i.e., grid cells). In the example, the grid may be regular and / or the cell may be cubic. The spatial location of the cell may correspond to the corresponding cell. In the example, two cells may not be located at the same spatial location. When simulating the evolution of a tumor, this spatial location allows one or more updated data segments to provide additional information about the spatial distribution of the degree of activation of an oncogene. In fact, when locating cells corresponding to multiple updated data segments, the spatial intra-tumor heterogeneity relative to the degree of activation of an oncogene becomes apparent. In the example, the simulation of tumor evolution may locate multiple updated cells in a three-dimensional or two-dimensional grid and provide the user with spatial intra-tumor heterogeneity of a simulated tumor.

[0054] In an example, the model provided at S20 may also be configured to determine whether the cell undergoes division. The determination may depend on the degree of activation of oncogenes within the cell. Then, depending on the determination performed by the model, when the model is run (S30), the information about cell proliferation may include information related to the presence or absence of cell division.

[0055] When the model determines that the corresponding cell divides, the information about the corresponding cell proliferation output at S30 can allow the spatial positioning of the two cells produced by the division to be determined. In addition, the information about the proliferation can further allow the degree of activation of the oncogenes in the two cells to be determined. In an example, the spatial positioning of the two cells produced by the division and the degree of activation of their corresponding oncogenes can be included in the information about the proliferation output at S30.

[0056] The spatial localization of the two cells produced by the division can be in the neighborhood of the parent cell. The neighborhood of a given cell can be the space of all points at a distance below a predetermined threshold. The distance can be a Euclidean distance. The predetermined threshold can be strictly lower than half of the maximum length of a two-dimensional or three-dimensional space. In the example, the predetermined threshold is lower than ten unit lengths and / or ten grid cells. In the example, the predetermined threshold is equal to one unit length and / or one grid cell. In other words, the neighborhood can correspond to all spatial localizations adjacent to a given cell. For example, two spatial localizations at a distance equal to the unit length and / or two spatial localizations corresponding to two grid cells sharing a common face, edge or vertex. The spatial localization of one of the two cells produced by the division can correspond to the spatial localization of the cell corresponding to the data segment used as input by the model. In the example, if there is no cell division, the spatial localization of the cell will not change during the update S40. Therefore, the spatial localization of the cell produced by the division can depend on the spatial localization of the cell undergoing division. Therefore, updating multiple data segments maintains the consistency of the initial distribution of the spatial localization of the cell. Therefore, the simulation can include the real spatial distribution of the next generation of cells and the activation degree of their corresponding oncogenes.

[0057] In an example, information about the proliferation of cells corresponding to a data segment used as input by the model may further depend on the spatial availability near the spatial location of the cell. Spatial availability refers to whether the cell is located at the spatial location. In other words, spatial availability depends on whether the spatial location has been assigned to any cell.

[0058] In an example, information related to the presence or absence of division of a cell as input can be based on spatial availability in the neighborhood of the cell. In an example, for the model, in order to positively determine during operation S30 that a cell corresponding to a data segment as input undergoes division, at least one spatial location near the spatial location of the cell must be an unoccupied spatial location and thus not assigned to any cell. Each unoccupied corresponding spatial location near the spatial location of the cell is an available spatial location that can be assigned to a cell produced by division. In an example, the model can determine which spatial locations are available based on the available data segments before running S30. Alternatively, the model can perform determinations during operation, allowing multiple simulations to run in parallel, each simulation corresponding to a different data segment.

[0059] In an example, if the model determines during operation S30 that the cell division corresponds to the data segment as input, the information about cell proliferation may include the corresponding activation degree of the oncogene for each cell produced by the division. The corresponding activation degree of the oncogene is based on the activation degree of the oncogene of the cell corresponding to the data segment used as input by the model. The activation degree of the oncogene of the cell produced by the division may be different from each other, and may also be different from the activation degree of the oncogene of the cell corresponding to the data segment used as input. In fact, during operation S30, the activation degree of the oncogene may change due to the simulated changes occurring in the cell. This allows the model to simulate the evolution of the tumor more closely. In an example, the model can take into account new changes, such as new mutations and / or asymmetric cell divisions. If the initial activation degree of the oncogene increases, the frequency of the new mutation and / or asymmetric cell division will increase. Alternatively or additionally, the activation degree of the oncogene of the cell produced by the division may be the same as the cell corresponding to the data segment used as input by the model. Therefore, the accuracy of the simulation of the evolution of the tumor can be improved. In effect, the approach can mimic tumors in which novel changes do not arise systemically in tumor cells.

[0060] In the example, the degree of activation of the oncogene of each cell produced by the division can correspond to a change in the degree of activation of the oncogene in the corresponding given cell that is at least partially probabilistic. "At least partially probabilistic" means that the change depends on a random variable. In addition, the change can further depend on a deterministic variable. In the example, the change in the degree of activation of the oncogene in the corresponding given cell can depend on a random variable and a fixed value, for example, a comparison of the generated value of the random variable with the fixed value. The random variable can be a number between 0 and 1 generated according to a uniform probability distribution on the range [0,1], and the fixed value can represent the probability of the occurrence of the biological process. The generated value of the random variable can then be compared with the fixed value, thereby simulating the probability in the iteration of running S30. Therefore, the model provided at S20 can take into account the probabilistic nature of the biological process that controls the evolution of the cell.

[0061] For example, the degree of activation of an oncogene in a cell can be increased based on the probability of a change. In an example, the probability of occurrence of one or more biological processes of a cell can be determined based on a threshold value determined according to a mathematical model. In addition, the calculations related to the changes in the degree of activation of oncogenes in a cell performed by the model during S20 can be performed only when the model determines that the cell undergoes division. Therefore, the model uses fewer computing resources during each run of S30 while maintaining biological considerations. In fact, changes may occur mainly during cell division. Additionally or alternatively, the degree of activation of oncogenes in cells produced by division can be based on the asymmetry of cell division that also follows probabilistic rules. Therefore, the evolution of the simulated tumor is based on the degree of activation of the oncogenes of the cell and is subject to the probabilistic changes inherent in the mechanism that controls cell evolution, which improves the accuracy of the simulation.

[0062] In an example, the model provided at S20 may be a decision-based model, such as a decision tree model. The decision-based model performs simulation by presenting decisions that result in at least two possible outcomes depending on the parameters of the model. In an example, the model may be such that once the outcome is determined, a new decision may be presented again. The new decision also results in at least two outcomes. The model provided at S20 may include a decision on the death or survival of a cell corresponding to the data segment as input. Then, if the result is cell survival, then a decision on the absence or presence of cell division may follow. Then, if the result is cell division, then a decision on obtaining a new change may follow, such as the degree of activation of an oncogene in a cell. The new change may be related to the acquisition of a new centromere, the acquisition of a new mutation, a change in a chromosome, a gain or loss of chromosome copy number, and / or asymmetric division. The decision may be affected by the degree of activation of an oncogene in a cell corresponding to the data segment as input. Alternatively, when a decision is made, the decision may be affected by the degree of activation of an oncogene in a cell. The decision-based model allows following the different stages of tumor cell evolution in the order in which they appear. The degree of activation of oncogenes in the resulting cells is therefore the result of a sequential simulation of biological processes. Thus, the accuracy of the simulation of the evolution of the tumor increases.

[0063] In addition, each decision can be determined based on the calculated results relative to the probability. The probability can depend on the degree of activation of the oncogene in the cell. In this way, the model provided at S20 simulates both the sequence of events that may lead to changes in the degree of activation of the oncogene in the cell and the probabilistic nature of the biological mechanism. Therefore, the accuracy of the simulation of the evolution of the tumor is further increased.

[0064] In an example, a treatment type may be provided, and information about proliferation may further depend on the treatment type. The treatment type may involve one or more treatments or a combination of treatments, such as drug therapy or therapeutic treatment. Then, the model provided at S20 may be further configured to consider the treatment type when determining whether the cell corresponding to the data segment as input survives. Determining whether the cell corresponding to the data segment as input survives may be performed at any time during operation. Therefore, the simulation of the evolution of the tumor may predict the relative efficacy of a specific treatment or treatment combination. In an example, the treatment type may be selective, such as a treatment targeting one or more specific oncogenes. Alternatively or additionally, the treatment type may be a general treatment type, such as chemotherapy. In this way, by considering the mechanism of action of one or more treatments, the accuracy of predicting the relative efficacy of a treatment or treatment combination may be improved.

[0065] In the example, after updating the plurality of data segments S40 with the results of the operation, iteration S50 can be performed by running the model again, taking one or more data segments from the updated plurality of data segments as input. The updated plurality of data segments may include new data segments generated by cell division in the previous operation of the model. For each iteration, after running the model S30 using one or more data segments from the updated plurality of data segments, the updated plurality of data segments are further updated based on the results of the new operation. This allows the evolution of a tumor over time to be simulated by simulating several generations of cells. Therefore, a simulation of the evolution of a tumor can be performed to estimate the tumor after a given amount of time has passed. In the example, each time a data segment is taken as input to the model, a time jump can be associated with the cell corresponding to the data segment.

[0066] In an example, after updating, the plurality of cells corresponding to the plurality of data segments are considered to be advanced in time by a duration equal to the time required to complete a cell cycle.

[0067] In the example, after the update, the plurality of data segments are considered to be advanced in time by a duration equal to the following:

[0068] (Time required to complete the cell cycle)

[0069] Where n is the number of data segments taken as input and N is the number of data segments in the plurality of data segments. For example, if one tenth of the data segments are taken as input, then after the update the cell population is advanced in time by one tenth of the time it takes to complete the cell cycle. This allows for asynchronous simulation of tumor cells.

[0070] In an example, before providing a model at S20, data related to a patient with a tumor may be provided. The model provided at S20 may then be adjusted based on the data related to the patient. "Adjustment" means that at least one parameter of the model may be modified according to the data related to the patient. If the parameter value of the model is empty, a value may be set. For example, the probability of a cell mutation may depend on lifestyle data or other clinical data of the patient. Lifestyle data may include any one or combination of smoking, body mass index (BMI) and / or age. Data related to the patient may include physical data. Physical data may include any one or combination of height, weight, body mass index (BMI) or any information related to the patient's lifestyle. Alternatively or additionally, data related to the patient may include the patient's medical information. The patient's medical information may include any one or combination of genetic risk factors, medical history, the patient's lifestyle, or any information related to the patient's genetic disease. Additionally or alternatively, the model may be adjusted by calibration to a specific patient population. In an example, specificity may be tumor location or oncogene.

[0071] By adjusting the model provided at S20 using data related to the patient, the simulation of the tumor evolution becomes personalized to the patient. This increases the accuracy of the simulated tumor evolution of the patient. Alternatively, the model provided at S20 can be configured according to data from the literature. For example, when data related to the patient is not available.

[0072] Reference now Figure 2 In an example, the model provided at S20 may include one or more parameters. The parameters may be determined by configuring values ​​so that the multiple data segments provided at S10 can be retrieved. The retrieval may be performed by defining one or more other data segments S22. Each other data segment may correspond to a given cell in an initial state of a tumor and include a degree of activation of an oncogene associated with a tumor in a given cell. The initial state of a tumor may include one or more cells that may lead to a tumor, such as a tumor represented by the multiple data segments provided at S10. One or more other data segments may correspond to a selected portion of the multiple data segments provided at S10. Alternatively or additionally, one or more other data segments may be assigned random spatial positioning and random activation degrees of oncogenes, respectively.

[0073] The parameters of the model are initialized S24. In other words, first values ​​of the model parameters are selected.

[0074] Next, the model is run S26 on one or more other data segments defined at S22.

[0075] Next, one or more other data segments are updated S27 based on the result of running S26.

[0076] Next, S26 and S27 may be iterated so as to increase the number of other data segments forming the plurality of other data segments and simulate the progression of the tumor in time.

[0077] Next, the updated plurality of other data segments are compared with the data segments of the plurality of data segments provided at S10 .

[0078] If the cells corresponding to the data segments of the updated plurality of other data segments are substantially the same as the cells corresponding to the data segments of the plurality of data segments provided at S10, the value of the parameter is verified S29. "Substantially the same" means that the cells corresponding to the other data segments of the updated plurality of other data segments and the cells corresponding to the data segments of the plurality of data segments provided at S10 have similar values, within a tolerance range that can be defined automatically or by user input. The similar values ​​between the cells can include the corresponding activation levels of oncogenes in the cells and / or the corresponding spatial localization of the cells. Alternatively, the comparison can involve the distribution of the activation levels of oncogenes between the cell populations.

[0079] If the cells corresponding to the other data segments of the updated multiple other data segments are not substantially the same as the cells corresponding to the data segments of the multiple data segments provided at S10, the parameter values ​​of the model are changed according to the similarity between the compared cells, and the process is repeated. Additionally or alternatively, several iterations of running and updating can be performed on the other data segments before comparing with the multiple data segments provided at S10. In the example, the iteration is performed until the number of other data segments of the updated multiple other data segments is substantially the same as the number of data segments of the multiple data segments provided at S10. Alternatively, the iteration can be performed so that the distribution of the activation degree of oncogenes in the cell population of the updated multiple other data segments is substantially the same as the distribution of the activation degree of oncogenes in the cells corresponding to the data segments provided at S10. The comparison between the cells corresponding to the other data segments of the updated multiple other data segments and the cells corresponding to the data segments of the multiple data segments provided at S10 can be performed between each iteration. Alternatively, the comparison can be performed after multiple iterations.

[0080] By adjusting the parameters of the model using the plurality of data segments provided at S10 as a comparison reference, the accuracy of the simulation of the evolution of a tumor performed using such a model is further increased. In fact, the cells corresponding to the data segments of the plurality of data segments provided at S10 represent real tumor cells. Therefore, retrieving real-world data from the simulation of a tumor sample indicates that the parameters of the model are correctly adjusted and will be implemented more accurately. Therefore, the simulation of the evolution of a tumor whose cells correspond to the data segments of the plurality of data segments provided at S10 will be accurate.

[0081] In the example, the first values ​​of the parameters of the model during initialization S24 are based on data related to the patient. In other words, the data related to the patient can be used as prior data for the model parameters. This increases the accuracy of the model for a given patient and avoids combinations of parameters that may not correspond to the patient. In fact, the model includes more than one parameter, so that there may be more than one combination of parameter values ​​that may result in similar simulations. Depending on the initial values ​​selected during the initialization of the parameters at S24, different combinations of parameter values ​​can be achieved when adjusting the model. By initializing the parameters based on the data related to the patient, the parameter values ​​verified at S29 are directly related to the patient's data. This further personalizes the simulation of the evolution of the tumor. Therefore, the accuracy of the simulation is increased.

[0082] Reference now Figure 3 , shows an example of a process of simulating the evolution of a tumor using calibration of a model. Past data from the patient's tumor is unknown because there is no clinical information about when it was first developed. To calibrate the model, a pool of cells is provided and a set of parameters are selected and / or calculated. Current data from the patient's tumor includes an image of the tumor from which a numerical representation of the current tumor can be calculated, for example in the form of data segments. The model is run using the pool of cells and the parameters of the past data until it is substantially the same as the numerical representation calculated from the current data. The model is then calibrated and can be used to perform numerical predictions of tumor evolution that represent future data.

[0083] Back to Figure 1 In an example, after the plurality of data segments are updated at least once at S40, statistics related to oncogenes and based on the plurality of data segments may be calculated. The statistics may include calculating the shape and value of the distribution of the activation levels of oncogenes in the cell population corresponding to the updated plurality of data segments. In addition, the statistics may further include the spatial distribution of the activation levels of oncogenes in the cells of the data segments corresponding to the updated plurality of data segments. Based on the statistics, a spatial representation of regions indicating different activation levels of oncogenes having a tumor may be further calculated. Based on the spatial representation or the calculated statistics, a user may retrieve predicted intra-tumor spatial heterogeneity. In addition, the statistics may include calculating a numerical value of the heterogeneity and presenting it to the user together with the tumor size.

[0084] The method for simulating the evolution of a tumor can be used to help medical experts choose the treatment that is most suitable for a patient. In fact, the simulation of the evolution of a tumor can predict the growth of the tumor, allowing medical experts to adjust the treatment priorities of the patient. Additionally or alternatively, assuming that the simulation can predict intra-tumor heterogeneity, the method can further help medical experts choose treatment options. In addition, the method is adapted to the time constraints of medical professionals. In fact, the simulation can focus on a limited number of oncogenes without performing calculations corresponding to a large part of the genome of the cell. This allows the simulation to be very short so that it can be used in a clinic or hospital for patients who cannot stay for a long time.

[0085] The method is computer-implemented. This means that the steps (or substantially all steps) of the method are performed by at least one computer or any similar system. Therefore, the steps of the method are performed by a computer, which may be fully automatic or semi-automatic. In an example, the triggering of at least some steps of the method may be performed by user-computer interaction. The required user-computer interaction level may depend on the foreseen level of automation and be balanced with the need to implement the user's wishes. In an example, the level may be user-defined and / or predefined.

[0086] A typical example of a computer implementation of the method is to perform the method using a system suitable for this purpose. The system may include a processor and a graphical user interface (GUI) coupled to a memory having a computer program recorded thereon, the computer program including instructions for performing the method. The memory may also store a database. The memory is any hardware suitable for such storage, possibly including several physically distinct parts (e.g., parts of the program and possibly parts of the database).

[0087] Fig.10 An example of the system is shown, where the system is a client computer system, such as a user's workstation.

[0088] The client computer of this example includes a central processing unit (CPU) 1010 connected to an internal communication bus 1000 and a random access memory (RAM) 1070 also connected to the bus. The client computer is also provided with a graphics processing unit (GPU) 1110, which is associated with a video random access memory 1100 connected to the bus. Video RAM 1100 is also known in the art as a frame buffer. A mass storage device controller 1020 manages access to a mass storage device such as a hard disk drive 1030. Mass storage devices suitable for tangibly embodying computer program instructions and data include all forms of non-volatile memory, including, for example, semiconductor memory devices such as EPROM, EEPROM, and flash memory devices; magnetic disks such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM disks 1040. Any of the foregoing may be supplemented by or incorporated into a specially designed ASIC (application specific integrated circuit). A network adapter 1050 manages access to a network 1060. The client computer may also include a tactile device 1090, such as a cursor control device, a keyboard, and the like. A cursor control device is used in the client computer to allow the user to selectively position the cursor at any desired location on the display 1080. In addition, the cursor control device allows the user to select various commands and input control signals. The cursor control device includes a plurality of signal generating devices for inputting control signals to the system. Typically, the cursor control device may be a mouse, the buttons of which are used to generate signals. Alternatively or additionally, the client computer system may include a sensitive pad and / or a sensitive screen.

[0089] The computer program may include instructions executable by a computer, the instructions including a module for causing the above-mentioned system to perform the method. The program may be recorded on any data storage medium (including the memory of the system). The program may be implemented, for example, in a digital electronic circuit, or in computer hardware, firmware, software, or a combination thereof. The program may be implemented as a device, such as a product tangibly embodied in a machine-readable storage device, for execution by a programmable processor. The method steps may be performed by a programmable processor executing an instruction program to perform the functions of the method by operating on input data and generating outputs. Therefore, the processor may be programmable and coupled to receive data and instructions from a data storage system, at least one input device, and at least one output device, and to send data and instructions to the data storage system, at least one input device, and at least one output device. If necessary, the application may be implemented using a high-level process-oriented or object-oriented programming language, or implemented using an assembly language or a machine language. In any case, the language may be a compiled language or an interpreted language. The program may be a complete installer or an updater. The application of the program on the system results in any case in the instructions for executing the method.

[0090] The following is a discussion of treatment decisions related to the tumors on which this approach is constructed.

[0091] This discussion refers to the following list of academic papers:

[0092] 1. Diagn Interv Imaging. 2014 Jun;95(6):527-39. Ammari S, Thiam R, Cuenod CA, et al. Radiological evaluation of response to treatment: application to metastatic renal cancers receiving anti-angiogenic treatment. Diagn Interv Imaging. 2014 Jun;95(6):527-39.

[0093] 2. Annals of Oncology, February 19, 2017. doi:10.1093 / annonc / mdx036. Cardoso F, et al. 3rd ESO-ESMO International Consensus Guidelines for Advanced Breast Cancer (ABC 3). Ann Oncol. 2017Feb 19. doi:10.1093 / annonc / mdx036.

[0094] 3. Journal of Molecular Diagnosis, 2013; 15(4): 415-53. Lindeman NI, et al. Molecular testing guideline for selection of lung cancer patients for EGFR and ALK tyrosine kinase inhibitors: guideline from the College of American Pathologists, International Association for the Study of Lung Cancer, and Association for Molecular Pathology. J Mol Diagn. 2013; 15(4): 415-53.

[0095] 4. Annals of Oncology, 2016 Sep; 27(suppl 5): v1-v27 asymptotic analysis. Computational & Applied Mathematics (2006). Novello S, et al. Metastatic non-small-cell lung cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2016 Sep; 27(suppl 5): v1-v27 asymptotic analysis. Computational & Applied Mathematics (2006).

[0096] 5. Watanabe T. Model Systems Facilitating an Understanding of Mechanisms for Oncogene Amplification. Oncogene and Cancer-From Bench to Clinic, book edited by Yahwardiah Siregar, ISBN 978-953-51-0858-0, published on January 24, 2013.

[0097] 6. Journal of Molecular Diagnostics, 2013, Wolff AC, et al. Recommendations for Human Epidermal Growth Factor Receptor 2. J Clin Oncol; 2013.

[0098] 7. Cell, 2006; 127(5): 905-915 suppl., "IEEE Transactions on Image Processing 2004" Anderson ARA, Weaver AM, Cummings PT, Quaranta V., "Tumor morphology and phenotypic evolution driven by selective pressure from the microenvironment. Cell. 2006; 127(5): 905–915 in painting." IEEE Transactions on Image Processing 2004.

[0099] 8. Scientific Reports, 2016;6:23383. Buckley, NE et al. Quantification of HER2 heterogeneity in breast cancer–implications for identification of sub-dominant clones for personalised treatment. Sci Rep. 2016;6:23383.

[0100] 9. Cottu PH, et al. Intratumoral heterogeneity of HER2 / neuexpression and its consequences for the management of advanced breast cancer. Ann Oncol 2008;19:595–7.

[0101] 10. Journal of Molecular Diagnostics, 2009;27(18):2962-9. Dowsett M et al. Disease-free survival according to degree of HER2 amplification for patients treated with adjuvant chemotherapy with or without 1year of trastuzumab: the HERA Trial. J Clin Oncol 2009;27(18):2962-9.

[0102] 11. PLoS One, 2017;12(9):e0184229, Iwasaki WM, et al. Simulation framework for generating intratumor heterogeneity patterns in a cancer cell population. PLoS One. 2017;12(9):e0184229.

[0103] 12. Lee HJ, et al. HER2 Heterogeneity Affects Trastuzumab Responses and Survival in Patients With HER2-Positive Metastatic Breast Cancer. Am J Clin Pathol 2014; 142: 755-766.

[0104] 13. Mathematical medicine and biology: a journal of the IMA, 34(2), 151-176. Lefebvre, G., Cornelis, F., Cumsille, P., Colin, T., Poignard, C., & Saut, O. (2016). Spatial modelling of tumour drug resistance: the case of GIST liver metastases. Mathematical medicine and biology: a journal of the IMA, 34(2), 151-176.

[0105] 14. Pathol Int. 2016;66(6):313-24. Nitta H et al. The assessment of HER2 status in breast cancer: the past, the present, and the future. Pathol Int. 2016;66(6):313-24.

[0106] 15. PLoS Comput. Biol. 2015;11(3):e1004025. Poleszczuk J, Hahnfeldt P, Enderling H. Evolution and phenotypic selection of cancer stem cells. PLoS Comput. Biol. 2015;11(3):e1004025.

[0107] 16. Ribba, B., Colin, T., Schnell, S., 2006. A multiscale mathematical model of cancer, and its use in analyzing irradiation therapies. Theoretical Biology and Medical Modelling 3, 1.

[0108] 17. American Journal of Clinical Pathology, 2012; 137: 595-605. Starczynski J et al. HER2 Gene Amplification in Breast Cancer: A Rogues' Gallery of Challenging Diagnostic Cases: UKNEQAS Interpretation Guidelines and Research Recommendations. Am J Clin Pathol 2012; 137: 595-605.

[0109] 18. Nature Genetics, 2015; 47(3): 209-216. Sottoriva A, Kang H, Ma Z, Graham TA, Salomon MP, Zhao J, et al. A Big Bang model of human colorectal tumor growth. Nat. Genet. 2015; 47(3): 209-216.

[0110] 19. Science Translational Medicine, 2012;4(127):127ps10, Yap TA, et al. Sci Transl Med. Intratumor heterogeneity: seeing the wood for the trees. 2012;4(127):127ps10.

[0111] 20. Nature Communications, 2016;7:12222. Ferrari A, et al. A whole-genome sequence and transcriptome perspective on HER2-positive breast cancers. Nat Commun. 2016;7:12222.

[0112] 21. Breast Cancer Research. 2012;14, R150. doi:10.1186 / bcr3362. Marotta M, Chen X, Inoshita A, Stephens R, Thomas Budd G, Crowe JP, et al. A common copy-number breakpoint of ERBB2 amplification in breast cancer colocalizes with a complex block of segmental duplications. Breast Cancer Research. 2012;14, R150. doi:10.1186 / bcr3362.

[0113] 22. Marotta M. Palindromic amplification of the ERBB2 oncogene in primary HER2-positive breast tumors. Sci Rep. 2017;7:41921.

[0114] 23. NICE guideline, accessed June 21, 2017, https: / / pathways.nice.org.uk / pathways / advanced-breast-cancer#path=view%3A / pathways / advanced-breast-cancer / managing-advanced-breast-cancer.xml&content=view-node%3Anodes-hrpos-and-her2pos(NICE guideline, accessed June 2017 21. https: / / pathways.nice.org.uk / pathways / advanced-breast-cancer#path=view%3A / pathways / advanced-breast-cancer / managing-advanced-breast-cancer.xml&content=view-node%3Anodes-hrpos-and-her2pos).

[0115] 24. NCCN guidelines, accessed 2017June 21. https: / / www.nccn.org / professionals / physician_gls / pdf / breast.pdf.

[0116] 25. Biochim Biophys Acta. 2015;1856(1):73-85. Barnard ME, et al. Established breast cancer risk factors and risk of intrinsic tumor subtypes. Biochim Biophys Acta. 2015;1856(1):73-85.

[0117] 26. Cancer Epidemiology, Biomarkers, and Prevention. 2014 Jan;23(1):84-97. O'Brien KM, et al. Breast cancer subtypes and previously established genetic risk factors: A Bayesian approach. Cancer Epidemiol Biomarkers Prev. 2014 Jan;23(1):84-97.

[0118] In medicine, treatment decisions are based on diagnosis. For example, an oncologist may treat invasive breast cancer after the tumor biopsy confirms the diagnosis and after extended radiological or pathological examinations. In precision medicine, biological markers (biomarkers) can be added to support clinical decisions because they help improve diagnosis and help predict treatment outcomes. For example, breast cancer is divided into different subgroups based on the expression of estrogen receptors, progesterone receptors, and HER2 (human epidermal receptor 2), and they guide the use of hormone therapy and anti-HER2 targeted therapy. Companion diagnostic tests are biomarkers specifically designed to be paired with specific drugs. Companion diagnostics are medical devices that can help doctors decide which treatment options to offer patients, as well as the dosage tailored to the patient. Currently, the Federal Drug Administration has approved 39 companion tests based on genetic or histological biomarkers.

[0119] A major limitation in the interpretation of biomarker (and companion diagnostic test) results is inter-individual variability as observed for many biological processes, meaning that values ​​are spread out relative to a statistical distribution. Another limitation in the availability of biomarkers is intra-tumor heterogeneity, a common phenomenon in which different populations of cancer cells can coexist in a single tumor. In this case, it is unclear which cell populations the companion test should target.

[0120] Current biomarkers do take into account inter-individual and intra-tumor heterogeneity, even when genetic intra-tumor heterogeneity is assessed by approved tests, such as, for example, protein expression by in situ hybridization (ISH) and / or immunohistochemistry (IHC) for detection of HER2ALK gene amplification, RET gene fusion, ROS1 gene fusion, CD274 (PD-L1) expression, ESR / PGR (hormone receptor) expression.

[0121] Currently, these biomarkers are based on discretization of quantitative measurements using unique cutoff points to identify eligible patients for specific treatments. Cutoff points for gene amplification or protein expression are derived from expert guidelines. Examples of cutoff points include:

[0122] - HER2 expression in more than 10% of cells (IHC)[6]

[0123] - HER2 gene amplification in more than 10% of cells (ISH), as determined by a mean HER2 copy number ≥

[0124] 6.0 signal / cell or HER2 / centromere ratio ≥ 2.0 defined [6]

[0125] - ALK gene fusion in more than 15% of cells[3].

[0126] Currently, all patients with a positive biomarker are eligible for FDA-approved drugs. For example, all patients with HER2-positive breast cancer are preferably treated with pertuzumab-trastuzumab plus chemotherapy in the first line and ado-trastuzumab (TDM-1) in the second line [2], regardless of HER2 pattern. In the example of lung cancer, patients with ALK-positive lung cancer are eligible to receive ALK inhibitors in the first line [4].

[0127] Although guidelines recommend clear cutoff points for interpretation, borderline cases or abnormal patterns of biomarkers often occur. Many studies have reported that approximately 10% of breast cancers have HER2 gene amplification heterogeneity [8,9,17], which is associated with worse survival when treated with trastuzumab, a therapeutic anti-HER2 monoclonal antibody

[12] .

[0128] For precision therapy, interpatient variability can also be taken into account. Indeed, Dowsett et al. showed that HER2 amplification levels follow a Gaussian distribution in a large patient population and demonstrated an inverse correlation between HER2 amplification levels and trastuzumab efficacy.

[0129] Cancer is a dynamic process. Treatment decisions in oncology are based on the effect of treatment and its evolution. One of the means to assess the dynamic evolution of tumors during treatment is to use the RECIST assessment criteria to compare tumor size on medical imaging before and after treatment. However, RECIST does not take into account hereditary tumor heterogeneity. In contrast, dynamic assessment of hereditary tumor heterogeneity biomarkers is not clinically feasible because follow-up of patients with multiple biopsies does not meet clinical constraints requiring rapid treatment decisions, treatment initiation, and ethical norms.

[0130] The proposed method for modeling the evolution of tumors takes into account heterogeneous cell populations and improves patient selection without using cutoff points or systematic rules, such as considering only the major population.

[0131] Methods for simulating tumor evolution can be used to authorize treatment decisions. For example, when genetic and / or histological biomarkers are required to predict the anti-tumor efficacy of different treatments or treatment combinations. The method for simulating tumor evolution can simulate the evolution of oncogenes during tumor growth and cancer treatment and their spatial intra-tumor heterogeneity. In the example of breast cancer cells, the oncogenes in the cells can be activated by different changes, such as increasing the number of gene copies, also known as amplification (i.e., HER2 amplification), activating mutations (such as KRAS p.G12C or EGFR p.L858R), or inter- or intra-chromosomal rearrangements that lead to gene fusions (i.e., ALK gene fusions).

[0132] This approach could improve predictions related to tumor growth and therapeutic efficacy based on the intratumor heterogeneity of oncogenes.

[0133] Now let's discuss how to implement this method.

[0134] In this implementation, a plurality of data segments are provided, each corresponding to a patient's tumor cells. Each data segment includes the degree of activation of one or more oncogenes associated with the tumor and the spatial location assigned to the cell. The degree of activation of each oncogene can be a numerical value corresponding to the number of oncogenes in the cell. The plurality of data segments can be determined based on an image of the patient's tumor cells. A model is also provided, which is configured to take an input data segment and output information about the proliferation of a corresponding cell corresponding to the input data segment.

[0135] Information about proliferation includes information about cell death or survival, and / or information about the presence or absence of cell division, and / or information about the acquisition of one or more new changes during division. Information about the acquisition of one or more new changes may include information about the acquisition of any one or combination of centromeres, breakage-fusion-bridge extension events, and / or unbalanced chromosome segregation.

[0136] The information about proliferation depends on the degree of activation of one or more oncogenes in the cell. Notably, as the degree of activation increases, the information about proliferation is more likely to indicate the presence of division. In addition, as the degree of activation increases, the information about proliferation is more likely to indicate the acquisition of one or more new changes.

[0137] Furthermore, information about proliferation indicates the presence of divisions, which depends on the availability of space in the vicinity of the cell. It is noteworthy that information about proliferation indicates the presence of divisions only if there is space availability in the vicinity of the cell.

[0138] When the information about proliferation includes information about the presence of a split, the corresponding spatial location of each cell resulting from the split is also included in the information about proliferation. Information about the corresponding spatial location of each cell resulting from the split allows for the assignment of a spatial location to each cell resulting from the split. The spatial location of each cell resulting from a single split can be near each other. The spatial location of one of the cells resulting from the split is located at the spatial location of the cell corresponding to the data segment used as input.

[0139] When the information about proliferation includes information about the presence of division, the information about proliferation also includes the corresponding degree of activation of one or more oncogenes in each cell produced by the division. The degree of activation of one or more oncogenes in each cell produced by the division depends on the degree of activation of one or more oncogenes of the cell corresponding to the data segment as input.

[0140] The model is run on one or more data segments. Multiple data segments are updated based on the results of the run. The update is performed based on the proliferation information of the cells corresponding to the data segments as input. Therefore, the update is cell-wise because information about a single cell is used to update multiple data segments.

[0141] When updating, when the information about the proliferation includes the presence of division, one or more data segments are created. Due to the spatial positioning included in the data segments before and after the operation, the cells corresponding to the multiple data segments are spatially coordinated. Therefore, the spatial arrangement of the cells corresponding to the multiple data segments follows the constraints of cell division.

[0142] At the time of updating, when the information on cell proliferation corresponding to the at least one data segment includes information related to cell death, at least one data segment is discarded.

[0143] When updating, when the information of cell proliferation corresponding to the at least one data segment includes information about the absence of division or at least does not include information about the presence of division, the activation degree of one or more oncogenes included in the at least one data segment remains unchanged.

[0144] When the information on the proliferation of at least one cell corresponding to the data segment as input includes obtaining at least one change at the time of updating, the degree of activation of one or more oncogenes in at least one cell produced by the division increases.

[0145] After updating the plurality of data segments, one or more iterations are repeated. The iteration includes running the model on one or more of the updated data segments and updating the updated data segments based on the results of the run. Any cell corresponding to a data segment as input corresponds to a cell at a later time point. Thus, over several iterations, the model simulates the evolution of the tumor. Due to the data segments corresponding to individual cells of the tumor, the accuracy of the simulation of tumor evolution is increased. The data injected into the model at each run is relatively small (i.e., consisting only of the degree of activation and spatial localization of one or more oncogenes).

[0146] In addition, a treatment type is provided. The treatment type includes any one or combination of targeted therapy and / or general therapy. Targeted therapy is more effective for cells with one or more specific oncogenes. In this case, the information about proliferation depends on the treatment type. When a treatment type effective for cells corresponding to multiple data segments is provided, the information about proliferation is more likely to include information about cell death. By considering the treatment type, the simulation of tumor evolution predicts the efficacy of the treatment type relative to the tumor. Different treatment types can be provided. The simulation of tumor evolution can provide a comparison of the efficacy of each treatment type.

[0147] The model comprises fixed parameters. Optionally, patient data is provided and the parameters of the model are adjusted according to the patient data. The patient data and the data according to which the plurality of data segments are determined are from the same patient. The parameter adjustment is performed by changing one or more values ​​of the parameters. The patient data is information related to the patient. The patient data comprises physical information and / or medical information. By adjusting the parameters from the patient data, the simulation of the tumor evolution takes into account additional data specific to the patient. Thus, the simulation of the tumor evolution is personalized to the patient and its accuracy is increased.

[0148] The following are references Figure 3-8 An example of such an implementation of the method for listing scholarly literature is provided previously.

[0149] The evolution of tumor heterogeneity over time can capture inter-individual and intra-tumor heterogeneity and can therefore help predict optimal treatment strategies, including combinations of treatments. In contrast, current biomarker development focuses on selecting an “optimal” cutoff point to select patient populations that will benefit from a specific treatment.

[0150] Furthermore, the proposed approach is based on a specific model for the biological drivers of oncogene amplification.

[0151] This approach can predict tumor evolution from a single time point analysis, which improves on RECIST evolution measures.

[0152] The model can be coded in any programming language, including but not limited to python and OpenGL. The method addresses personalized medicine and can take into account genomics, patient and tumor characteristics (such as type and / or location) as well as patient characteristics.

[0153] refer to Figure 4 , multiple data segments can be calculated based on single cell data including information for oncogene activation. Single cell data for oncogene activation (i.e., HER2 amplification) can be obtained by FISH image analysis by counting the number of HER2 and CEP17 signals. Other types of images of tumor cells can also be used to extract tumor cell data. Figure 4 FISH analysis published by Starczynski J et al. is shown. FISH images can be processed to depict each cell nucleus, quantifying CEP17 and HER2 signals in each cell to provide a distribution of HER2 copy number heterogeneity. 1000 data segments were created, which represent an initial tumor cell population similar to the cell population of the analyzed image. In other words, multiple data segments corresponding to tumor cells are provided from the extracted tumor cell data, each tumor cell having a degree of activation of HER2. In addition, each data segment is attributed to a spatial localization in a three-dimensional space representing the tumor.

[0154] Figure 5A and Figure 5B The corresponding parts of the examples of the models are shown. Figure 5A The portion shown is connected to the Figure 5B The model is based on the biological knowledge of genetic mechanisms of oncogene activation (such as gene amplification, fusion, mutation). The model is also based on the functional effects of oncogenes on core cellular pathways (such as cell growth, cell division, apoptosis, paracrine secretion). For HER2 gene amplification, the model takes into account:

[0155] -Cell division rate, which depends on oncogene status (i.e., HER2 amplification level);

[0156] -Cell death rate, which may depend on the type of treatment (such as chemotherapy and anti-HER2 antibodies);

[0157] - Probability of a break-fusion-bridge event [20, 21, 25, 5], (such as a break at the telomeric HER2 gene), leading to bicentric chromosome 17 and initiating HER2 amplification (initiating event);

[0158] -Probability of break-fusion-bridge only when dicentric chromosome 17 is present (elongation event);

[0159] -The probability of unbalanced mitotic chromosome segregation, resulting in asymmetric divisions due to an increase in dicentric chromosomes.

[0160] The model is configured to simulate the evolution of a tumor through asynchronous cellular automation. Individual cells are updated independently so that the new state of a cell affects the computation of states in neighboring cells, and a decision tree model is used to stochastically drive cellular events.

[0161] Reference now Figure 6 ,The patient’s clinical and genomic data are used to identify the model parameter values ​​in a patient-individualized manner. Figure 6 An example of a GUI on which patient-related data may be entered is shown. The patient-related data may further be stored and retrieved from a database. Based on the patient-related data, the values ​​of the model parameters are adjusted.

[0162] In this example, the model further considers three treatment types for advanced HER2 breast cancer: chemotherapy (such as paclitaxel), anti-HER2 therapy (such as TDM-1 Trastuzumab-Ado-Emtensine) and a combination of chemotherapy and anti-HER2 (such as Trastuzumab + Pertuzumab + Paclitaxel). These three treatment types can be modeled according to their mechanism of action. Each treatment type can be calculated as a vector that returns a value to each mechanism of action included in the treatment type.

[0163] Each treatment has its own mode of action. Chemotherapy is a non-selective treatment. It kills cells independently of the state and nature of the cells. Chemotherapy can be modeled by the following probability function:

[0164]

[0165] where chimio_eff is the efficacy of chemotherapy.

[0166] If there are more copies of HER2 in the cells, then anti-HER2 therapy is a more effective targeted therapy. This phenomenon can be modeled by a probability function, such as the following Death_AntiHER probability function:

[0167]

[0168] in:

[0169] -β is a positive parameter;

[0170] -#HER is the degree of activation of HER2 and depends on the number of HER2 copies. If #HER2 is below 2, the probability of death is zero.

[0171] Alternatively, a probability function known in the art may be selected.

[0172] Back to Figure 5A , once the model is initialized, cells corresponding to data segments are randomly selected and at each decision a decision tree is applied with a random value “r” linearly distributed between 0 and 1. The value of r is compared to the probability associated with the decision and the outcome is selected accordingly.

[0173] The cell goes through different decision steps. First, a decision is calculated about whether the cell should survive based on the effect of the treatment that induces cell death. If the cell dies, a new cell is randomly selected. If the cell survives, a new decision is taken about whether the cell should divide.

[0174] The probability of a cell dividing depends on the cell's growth rate, which can be modeled using 3 parameters determined by a combination of a covariation model, prior biological knowledge, and replay of tumor history. An exemplary probability function that can model growth rate is shown below:

[0175]

[0176]

[0177] in:

[0178] -γ 0 is between 0 and 1 and is the reference growth rate;

[0179] -α is a parameter that regulates growth rate according to the strength of #HER2 and is patient dependent;

[0180] -#HER2 is the degree of HER2 activation.

[0181] Furthermore, a verification about the availability of space is performed in the vicinity of the range R from the spatial localization of the cell. The parameter R can be fixed during the initialization of the model and corresponds to the distance to the spatial localization of the cell. The cell divides only if the random draw r is below the value given by the GrowthRate function and there is an available spatial localization for the new cell resulting from the division. In this case, the path of arrow A501 is followed and at Figure 5B Continue on. The spatial positioning of the two cells that will result from cell division can be determined at this step. If either of the two conditions for cell division is not met, there is no division and a new cell is randomly selected.

[0182] Reference now Figure 5B, after determining that the cell has undergone division, follow the path of arrow A501. Next, a verification is performed on the degree of activation of HER2 within the cell. The verification may include whether each chromosome of the cell has exactly 1 centromere and 1 copy of HER2. If this is the case, the probability that the cell has an initiation event (double-stranded DNA break followed by a break-fusion-bridge (BFB)) around HER2 is calculated. This is also known as gaining a centromere. The probability of gaining a centromere can be modeled by a function called P_gain_centromere. This probability function can be modeled according to a mixed nonlinear model with patient-related data as covariates, such as BMI index, age

[25] and smoking and genetic risk factors

[26] ; such as mutations on FGFR2:rs2981579 and FGFR2:2981582. An example of the logarithm of the function is given below:

[0183] Log(P_gain_centromere j ) = Log(P_gain_ceontromere basal )+p×COV j +η j

[0184] in

[0185] - The average probability of obtaining a centromere.

[0186] - Weight parameter;

[0187] - Considering the BMI index, the patient's age and the covariance matrix of the two genetic risk factors FGFR2:rs2981579 and FGFR2:2981582;

[0188] - Among them, ω 2 (≈0.01) takes into account the random noise of the variability of biological mechanisms.

[0189] If the cell already has more than one centromere or more than one copy of HER2 on at least one chromosome, then instead of calculating the probability of obtaining a priming event around HER2, the probability of a dicentric cell having a BFB extension event is calculated. The probability of a dicentric cell having a BFB extension event can be modeled similarly to the probability of obtaining a priming event around HER2. The calculation can be performed by comparing the result of the random value r to the modeled probability.

[0190] Next, calculations were performed to determine whether the cells had gained extra chromosomes, in addition to the duplicated chromosomes that occur during cell division. If the cells had gained extra chromosomes, the activation of HER2 increased significantly.

[0191] Next, the probability of a cell having an unbalanced chromosome segregation, resulting in an asymmetric division, was calculated. Normal cell division is symmetrical, in which case the two cells resulting from the division have the same degree of HER2 activation. However, cells can divide asymmetrically, resulting in a difference between the degree of HER2 activation between the two cells resulting from the division. Figure 5B In the example shown, the first cell resulting from the asymmetric cell division has two centromeres and three copies of HER2 per chromosome. The second cell resulting from the asymmetric cell division has only one copy of HER2 per chromosome and one centromere. The determination of the presence or absence of asymmetric cell division can be calculated similarly to the determination of obtaining additional chromosomes. If the cell undergoes asymmetric cell division, a second calculation can determine whether the asymmetric cell division is strong. Strong asymmetric cell division is asymmetric cell division in which the number of chromosomes between the two cells resulting from the division is unbalanced. In Figure 5B In the example of , the first cell resulting from strong asymmetric cell division has three chromosomes, while the second cell has only one chromosome. Calculations to determine the presence of strong asymmetric cell division can be performed similarly to the presence of asymmetric cell division.

[0192] After the cell splits, two new cells are created. Each cell corresponds to a corresponding data segment. One of the corresponding data segments may replace the data segment corresponding to the cell that has undergone the decision tree. Another data segment may be created after the run. The multiple data segments may be updated with the results generated by the decision tree after each run. Alternatively, the multiple data segments may be updated after all cells corresponding to the multiple data segments have passed through the decision tree. The determination of space availability may take into account new cells generated by cell splitting, even if the multiple data segments have not been updated.

[0193] After all cells have passed through the decision tree, a new iteration of the model can be performed with the updated multiple data segments.

[0194] Figure 7Examples of spatial representations of cells corresponding to multiple data segments before passing through the model and after one or more runs of the model are shown. By differences in intensity and color, the degree of activation of HER2 in the cell can also be represented. Alternatively, other methods of distinguishing the degree of activation of oncogenes in the cell can be used when calculating the representation of the cell in its corresponding spatial localization. The spatial representation of the cell after passing through the model is a simulation of how the tumor evolves from its initial state to the final state. The final state depends on the data extracted from the cells used to create the initial multiple data segments, as well as data related to the patient and / or treatment type. The spatial representation can be calculated after each run of the model, so that the real-time evolution of the tumor can be visualized by the user. This visual representation can be directly interpreted by a pathologist for histological analysis.

[0195] Reference now Figure 8 Alternatively or additionally, a copy number graph can be calculated based on the updated multiple data segments, as if the tumor had been sequenced and then analyzed. The graph can correspond to the entire updated multiple data segments. Alternatively, the graph can correspond to a portion of the multiple data segments, such as a data segment corresponding to a cell located in a two-dimensional plane passing through the simulated tumor. Confidence intervals for the average copy number and noise distribution between cells can be further calculated. A histogram of the HER2 copy number distribution in the total cell population (such as Figure 8 ) can be further presented to the user. Computing such statistics aids geneticists in interpreting the results, as the genetic information extracted from the simulation is presented in a format similar to that of genomics analysis.

[0196] Reference now Fig. 9 Alternatively or additionally, the percentage of tumor evolution can be calculated based on the following formula:

[0197]

[0198] This calculation helps to compare tumor evolution with other similar cases. In addition, several simulations for the same tumor can be presented to the user at the same time. For example, simulations produced by different treatment combinations, each showing the percentage of tumor evolution when compared with the initial tumor size (dashed line). In addition, according to the degree of activation of HER2 in the cell, the comparative value of intratumor heterogeneity can be calculated and presented to the user. These calculations help oncologists to interpret the simulation results because they are similar to the RECIST criteria. The RECIST criteria only allow tracking the evolution of tumor volume, but do not provide information about tumor heterogeneity. The method also provides information about the evolution of tumor heterogeneity.

[0199] Additionally, the method for simulating tumor evolution can be performed using the degree of activation of HER2 and the degree of activation of one or more other oncogenes that are also present in breast cancer. In this case, the parameter values ​​of the model can depend on the one or more oncogenes, similar to how the parameters depend on HER2 in the above example.

[0200] Fig.11 An example of a data flow of a model example of a method for model construction and for simulating tumor evolution. Based on biological knowledge (100) of tumor-related phenomena, a model (320) can be constructed (S310). Parameters and initial conditions are injected into the model and data is output. The in parameters and / or initial conditions can be input data of the model at a given time t. The output data can correspond to the time evolution of the input data, i.e., the input data at a future time t+t'.

[0201] Fig.12 is an example of a model that outputs a prediction of the time evolution of input data (900) for simulating tumor evolution. The input data (510) represents a cell pool, for example, the input data can be a data segment, each data segment represents a tumor cell. Parameters (500) are injected into the model for simulation. The model is run (S300). The run can include iterating (S340) the input-output from the model (320) several times. At each iteration, the model outputs a prediction (S330) of the input data for a future time. For example, after each iteration, the time evolution of the cell pool can be advanced in time by a duration equal to the cell cycle. Alternatively, depending on the input data, the cell pool can be advanced in time by a duration equal to a small portion of the cell cycle. After several iterations, a prediction of the time evolution of the input data can be calculated.

[0202] Fig.13is an example illustrating the adjustment of parameters (patient parameters) of a model for a given patient. Patient parameters (500) are adjusted based on a combination or any one of patient-related data (120), group data (130), biological knowledge (100), and tumor data (110). In other examples, patient parameters may be determined using any one or a combination of patient-related data, group data, biological knowledge, and tumor data. The group data may be data from other patients with similar tumors to the patient. Based on the group data, a covariation model (S600) may be established, which is known in the literature. A covariation model (S610) may be run on the patient-related data (120), which may determine patient parameters. Patient parameters may also be adjusted based on biological knowledge (100), for example, data from the literature. Group data (130) and biological knowledge (100) may be combined to create average model parameters (530). These average model parameters represent parameters of the model that are adjusted for the average member of the population. The average model parameters are injected into the model adjustment (700). Tumor data (100), such as data extracted from a biological image of a patient's tumor, can be used to compare output data from a model run in which the model was parameterized with the average model data in order to adjust the model for the patient as previously described (S700). The patient parameters (500) can be further adjusted by running the model. The model can then output numerical data representing the patient's tumor (510).

[0203] Fig.14 is an example of model adjustment at S700. In model adjustment, random noise can be applied (S710) to the average parameters (530) before the average parameters (530) are injected into the model. Random initial tumor data (S720) is calculated as input data and the model is run (S300). A digital tumor evolution (900) corresponding to the selection of the parameters and input data is output. The model can be run again multiple times on the output data before being compared with the patient's tumor data (such as data extracted from the patient's tumor image). The calculation of random initial tumor data and the application of random noise to the average parameters and subsequent model runs are repeated (S730) until the digital tumor fully represents the real tumor data. The model parameters of the numerical tumor calculated from which the real tumor data is fully represented can then become the patient parameters (500), and the model can represent the numerical data (510) of the patient's tumor because the parameters are now personalized to the patient.

Claims

1. A computer-implemented method for simulating the evolution of a tumor associated with an oncogene, the method comprising: include: Providing (S10) a plurality of data segments, each data segment corresponding to a given cell of the tumor, each data segment comprising a degree of activation of the oncogene in the given cell; providing (S20) a model, the model being configured to acquire an input data segment and output information about proliferation of a corresponding given cell corresponding to the input data segment, the information about proliferation depending on the degree of activation of the oncogene in the corresponding given cell; running (S30) the model on one or more data segments of the plurality of data segments; and updating (S40) the plurality of data segments based on the result of the operation (S30); wherein each data segment further comprises the spatial location of said given cell of said tumor; in: The information about the proliferation of the corresponding given cell comprises information about the presence or absence of division of the corresponding given cell, and When the information about the proliferation of the corresponding given cell includes information about the presence of division, the information about the proliferation of the corresponding given cell also includes information about the corresponding spatial location for each cell produced by the division, each corresponding spatial location being near the spatial location of the corresponding given cell; wherein the information about the proliferation of the corresponding given cell also depends on the availability of space in the vicinity of the corresponding given cell; in: The information about the presence or absence of division of the respective given cell is based on the availability of the space in the vicinity of the respective given cell, and When the information about the proliferation of the corresponding given cell includes information about the presence of division, the corresponding spatial location near the spatial location of the corresponding given cell is an unoccupied spatial location; wherein, when the information about the proliferation of the corresponding given cell includes information related to the presence of division, the information about the proliferation of the corresponding given cell also includes the corresponding activation degree of the oncogene for each cell generated by the division based on the activation degree of the oncogene of the corresponding given cell; wherein the respective degree of activation of an oncogene in each cell resulting from said division corresponds to an at least partially probabilistic change in the degree of activation of said oncogene in said respective given cell.

2. The method according to claim 1, in, The model is decision-based and includes the following with respect to the corresponding given cell: Decisions about dying or living, Decisions regarding the presence or absence of a split, and A decision about acquiring one or more new changes during a split.

3. The method according to claim 1 or 2, in, The method further comprises providing a type of treatment, the information about the proliferation being dependent on the type of treatment.

4. The method according to claim 1 or 2, in, The providing, running and updating are iterated ( S50 ), and the plurality of data segments provided in each iteration are the updated plurality of data segments of the previous iteration.

5. The method according to claim 1 or 2, further comprising providing data related to a patient suffering from the tumor, the model comprising parameters depending on the data related to the patient.

6. The method according to claim 5, in, Providing the model includes determining the parameter as a value configured to retrieve the plurality of data segments by: defining one or more other data segments, each other data segment corresponding to a given cell in the initial state of the tumor, each other data segment comprising the activation degree of the oncogene in the given cell; running the model on one or more other data segments; and updating the plurality of other data segments based on the results of the operation, Wherein, optionally, the value of the parameter is initialized based on data related to the patient.

7. The method according to claim 1 or 2, further comprising determining statistical data related to the oncogene based on the updated plurality of data segments.

8. A method for providing a data structure, the data structure comprising a model, the model being configured to perform the method according to any one of claims 1 to 7.

9. A computer program product having a computer program stored thereon, which, when executed by a processor, causes the processor to perform the method according to any one of claims 1 to 7.

10. A system comprising a processor coupled to a memory having recorded thereon the computer program according to claim 9.

Images