Spatial stochastic models of cancer: Fitness, migration, invasion

  • Cancer progression is driven by genetic and epigenetic events giving rise to heterogeneity of cellphenotypes, and by selection forces that shape the changingcomposition of tumors. The selection forces are dynamic and depend onmany factors. The cells favored by selection are said to be more``fit'' than others. They tend to leave more viable offspring andspread through the population. What cellular characteristics makecertain cells more fit than others? What combinations of the mutantcharacteristics and ``background'' characteristics make the mutantcells win the evolutionary competition? In this review we concentrateon two phenotypic characteristics of cells: their reproductivepotential and their motility. We show that migration has a directpositive impact on the ability of a single mutant cell to invade apre-existing colony. Thus, a decrease in the reproductive potentialcan be compensated by an increase in cell migration. We furtherdemonstrate that the neutral ridges (the set of all types with theinvasion probability equal to that of the host cells) remain invariantunder the increase of system size (for large system sizes), thusmaking the invasion probability a universal characteristic of thecells' selection status. We list very general conditions underwhich the optimal phenotype is just one single strategy (thus leadingto a nearly-homogeneous type invading the colony), or a large set ofstrategies that differ by their reproductive potentials and migrationcharacteristics, but have a nearly-equal fitness. In the latter casethe evolutionary competition will result in a highly heterogeneouspopulation.

    Citation: Natalia L. Komarova. Spatial stochastic models of cancer: Fitness, migration, invasion[J]. Mathematical Biosciences and Engineering, 2013, 10(3): 761-775. doi: 10.3934/mbe.2013.10.761

    Related Papers:

    [1] Chunxiang Fan, Zouqin Huang, Binbin Chen, Baojin Chen, Qi Wang, Weidong Liu, Donghai Yu . Comprehensive analysis of key lncRNAs in ischemic stroke. Mathematical Biosciences and Engineering, 2020, 17(2): 1318-1328. doi: 10.3934/mbe.2020066
    [2] Jie Wang, Md. Nazim Uddin, Rehana Akter, Yun Wu . Contribution of endothelial cell-derived transcriptomes to the colon cancer based on bioinformatics analysis. Mathematical Biosciences and Engineering, 2021, 18(6): 7280-7300. doi: 10.3934/mbe.2021360
    [3] Xuesi Chen, Qijun Zhang, Qin Zhang . Predicting potential biomarkers and immune infiltration characteristics in heart failure. Mathematical Biosciences and Engineering, 2022, 19(9): 8671-8688. doi: 10.3934/mbe.2022402
    [4] Chenyang Jiang, Weidong Jiang . AGTR1, PLTP, and SCG2 associated with immune genes and immune cell infiltration in calcific aortic valve stenosis: analysis from integrated bioinformatics and machine learning. Mathematical Biosciences and Engineering, 2022, 19(4): 3787-3802. doi: 10.3934/mbe.2022174
    [5] Jun Wang, Mingzhi Gong, Zhenggang Xiong, Yangyang Zhao, Deguo Xing . Immune-related prognostic genes signatures in the tumor microenvironment of sarcoma. Mathematical Biosciences and Engineering, 2021, 18(3): 2243-2257. doi: 10.3934/mbe.2021113
    [6] Yong Luo, Qingbiao Chen, Jingbo Lin . Identification and validation of a tumor mutation burden-related signature combined with immune microenvironment infiltration in adrenocortical carcinoma. Mathematical Biosciences and Engineering, 2022, 19(7): 7055-7075. doi: 10.3934/mbe.2022333
    [7] Jianpei Hu, Zengnan Mo . Dissection of tumor antigens and immune landscape in clear cell renal cell carcinoma: Preconditions for development and precision medicine of mRNA vaccine. Mathematical Biosciences and Engineering, 2023, 20(2): 2157-2182. doi: 10.3934/mbe.2023100
    [8] Yuedan Wang, Jinke Huang, Jiaqi Zhang, Fengyun Wang, Xudong Tang . Identifying biomarkers associated with the diagnosis of ulcerative colitis via bioinformatics and machine learning. Mathematical Biosciences and Engineering, 2023, 20(6): 10741-10756. doi: 10.3934/mbe.2023476
    [9] Dong-feng Li, Aisikeer Tulahong, Md. Nazim Uddin, Huan Zhao, Hua Zhang . Meta-analysis identifying epithelial-derived transcriptomes predicts poor clinical outcome and immune infiltrations in ovarian cancer. Mathematical Biosciences and Engineering, 2021, 18(5): 6527-6551. doi: 10.3934/mbe.2021324
    [10] Xiaodong Xia, Manman Wang, Jiao Li, Qiang Chen, Heng Jin, Xue Liang, Lijun Wang . Identification of potential genes associated with immune cell infiltration in atherosclerosis. Mathematical Biosciences and Engineering, 2021, 18(3): 2230-2242. doi: 10.3934/mbe.2021112
  • Cancer progression is driven by genetic and epigenetic events giving rise to heterogeneity of cellphenotypes, and by selection forces that shape the changingcomposition of tumors. The selection forces are dynamic and depend onmany factors. The cells favored by selection are said to be more``fit'' than others. They tend to leave more viable offspring andspread through the population. What cellular characteristics makecertain cells more fit than others? What combinations of the mutantcharacteristics and ``background'' characteristics make the mutantcells win the evolutionary competition? In this review we concentrateon two phenotypic characteristics of cells: their reproductivepotential and their motility. We show that migration has a directpositive impact on the ability of a single mutant cell to invade apre-existing colony. Thus, a decrease in the reproductive potentialcan be compensated by an increase in cell migration. We furtherdemonstrate that the neutral ridges (the set of all types with theinvasion probability equal to that of the host cells) remain invariantunder the increase of system size (for large system sizes), thusmaking the invasion probability a universal characteristic of thecells' selection status. We list very general conditions underwhich the optimal phenotype is just one single strategy (thus leadingto a nearly-homogeneous type invading the colony), or a large set ofstrategies that differ by their reproductive potentials and migrationcharacteristics, but have a nearly-equal fitness. In the latter casethe evolutionary competition will result in a highly heterogeneouspopulation.


    1. Introduction

    Polluted sites, no matter if we are about talking the terrestrial or aquatic environment are omnipresent and significant efforts to disclose the sites and subsequently to clean up are made. However, remediation is not necessarily easy and not always straight forward. Thus, possibly a variety of remediation technologies may be brought into play, the different method having their advantages and disadvantages. Eventually, the choice of technology will be based on a weighing on pros and cons for the single options.

    Typically, the choice cannot be based on a single indicator expressing the effectiveness of a given remediation approach. Thus, in order to rank the different possible remediation approaches in question requires several indicators as a proxy to an appropriate ranking. In other words, we are dealing with a so-called multi-indicator system (MIS) [1]. Based on such a set of indicators it is possible not only mutually to rank the single remediation approaches but also to obtain some ideas to what extent one approach is better, i.e., more efficient than another.

    An alternative to handling the MIS with an appropriate mathematical model is, what is often seen a mathematical mapping of the single indicator values to get a one-dimensional scalar, eventually to be used as the ranking indicator [2]. However, this is not advisable as such a mapping process not only hides all the background information but also may have an unwanted effect of compensation [3].

    There are many well-known methods to obtain from a multivariate data matrix a linear (or with respect to technical aspects also a weak, i.e., including ties) order like PROMETHEE [4], or the ELECTRE family [5, 6, 7, 8, 9]. However, all these methods require parameters beyond the data matrix to run them and may even be criticized from a theoretical point of view [3]. In the present paper, the concepts of partial order methodology are applied to the evaluation of remediation technologies. Partial ordering is, from a mathematical point of view simple, applying only the relation “≤” and appears as an advantageous way to look at MIS [10].

    2. Methods

    2.1. Partial ordering

    Bruggemann and Carlsen (2012) recently argue that a partial order methodology is an alternative approach to study multi-indicator system (MIS). Partial ordering is a non-parametric method as, in contrast to standard multidimensional analyses, such as PROMETHEE [4] or the ELECTRE family [5, 6, 7, 8, 9] as no assumptions about linearity or distribution of the indicators neither are made and nor are necessary. ELECTRE III vs partial order based methodologies has been reported by Bruggemann and Carlsen(2015) [11].

    Partial ordering has been described intensively in recent year [12, 13]. Hence, in the following only necessary basic information is given.

    2.1.1. Basics of partial ordering

    In Partial Ordering the only ma­ thematical relation among the objects, i.e., in the present case remediation options is “≤”. The “≤”-relation is the basis for a comparison of the different remediation options, ROs, and constitutes a graph. A given ROx is connected with ROy if and only if the relation y ≤ x holds. Thus, it needs to be defined what is meant by x ≤ y. If a system that can be described by a series of indicators rj, is considered, a ROx is characterized by a set of indicators rj(ROx), j = 1, . . ., m and can be compared to ROy, characterized by the indicators rj(ROy), when

    ri(ROy) ≤ ri(ROx) for all i = 1, …, m equation (1)

    In the present study m equals 3 (cf. section 2.2).

    Equation (1) is obviously a rather strict requirement for establishing a comparison demanding all indicators of ROx should be better (or at least equal) than the corresponding indicators of ROy. Note that in the present example (cf. the data section 2.2), the higher the indicator values are the better they are considered to be. Hence, if X is the group of ROs studied, i.e., X = {RO1, RO2, …, ROn}, ROx will be ordered higher (better) than ROy, i.e., ROx > ROy, if at least one of the indicator values for ROx is higher (better) than the corresponding indicator value for ROy and no indicator for ROx is lower (worse) than the corresponding indicator value for ROy. On the other hand, if for some indicator j rj(ROx) > rj(ROy) and for some other indicator i ri(ROx) < ri(ROy), ROx and ROy will be denoted incomparable (notation: ROx ǁ ROy). In mathematical term this means a contradiction due to conflicting indicator values prevails. In cases where all indicator values for two RO are equal for all j, the two options will be considered as equivalent, i.e., ROx ~ ROy, having the same order. Note that in other studies applying partial order methodology the term “order” is typically referred to as “rank” or “height”.

    The graphical representation of the partial ordering is often given in a so-called Hasse diagram [12], which is a directed graphs, which is structured in levels and consist of chains and antichains (Figure 1).

    Figure 1. Hasse diagram of the MIS given in Table l.


    · Chain: A subset X' ⊆ X, where all objects fulfill (1) is called a chain. A chain has a length, which is |X'|-1.For objects within a chain, e.g., from the bottom to the top of the chain, all indicators are simultaneously non-decreasing.

    · Antichain: A subset X' ⊆ X, where no object fulfill (1), i.e., all objects in X' are mutually incomparable, is called an antichain. Thus, for any two objects within an antichain there is a conflict in indicator values.

    · Level: The horizontal arrangement of objects within a Hasse diagram. Levels per definition constitute an antichain, whereas the reverse not necessarily is true.

    2.1.2.Average orders

    The level structure constitutes a first approximation to ordering. However, this will obviously give rise to many tied orders as all objects in a level automatically will be assigned identical orders. It is in general desirable that the degree of tiedness is as low as possible. Hence, ultimately a linear ordering of the single objects (here ROs) is desirable, which it is obviously not immediately obtainable when incomparable objects are included.

    Partial order methodology provides a weak order, where tied orders are not excluded. This is obtained by calculating the average order of the single objects as described by Bruggemann and Annoni (2014) [14]. In cases comprising only a relatively low number of objects (typically < 25) the average orders may be calculated by an exact method based on lattice theory [15, 16, 17]. For larger systems, approximations have been presented [1].

    The calculations, which in the present study the average orders are calculated based on the exact method will assign an average order to the individual objects [15, 16, 17]. With the indicator orientation chosen as in the present study, the average orders will be assigned from 1 (bottom) up to maximum n (top), n being the number of RO (here 6 (7), cf. section 2.2), which can be regarded as a scoring system, i.e., the higher the score the better the RO.

    2.1.3.Perturbation of order characteristics

    The single objects in the Hasse diagram are characterized by order characteristics, which are 4 values that state the number of equivalences, predecessors (number of upwards comparable objects), successors (number of downwards comparable objects) and incomparable elements, respectively [13]. Obviously, changes in one or more of the indicators for a given object may cause changes in the order characteristics for the object under investigation and thus be a tool to disclose where specific ROs advantageously could be improved [18].

    2.1.4.PyHasse software

    Partial order analyses were carried out applying the PyHasse software [19]. PyHasse is programmed using the interpreter language Python (version 2.6). The software package contains today more than 100 modules and is under continuous further development 1. The present study will apply only three central modules of the PyHasse suite, i.e., mHDCl2_8 (PyHasse main module), avrank5 (exact method to generate average orders) and the scanincomp4_4 (setting threshold limits for significance).

    1 PyHasse is available upon request from the developer, Dr. R. Bruggemann (brg_home@web.de). An Internet version is currently under development and will contain a limited number of the more frequently used module. The internet version that currently comprise a few central modules is available at www.PyHasse.org.

    2.2.Datasets

    To illustrate the principles of using partial order methodology as an analytical decision support tool for the selection of the most appropriate remediation technology for polluted sites a simple, but exemplary dataset has been created assuming 5 possible options for remediation and 3 polluting chemicals for which the remediation is supposed to remove to some extent. In Table 1 the dataset applied is shown. Note that in addition to the 5 remediation options a reference option, RO0, is introduced corresponding to the original concentrations, i.e., the remediation percentage for all 3 chemicals is 0.A further reference point, ROmax, is introduced referring to the optimal situation, i.e., all 3 chemicals being completely removed corresponding to a remediation percentage of 100%.

    Table 1. Fraction of three chemicals being removed by different remediation options
    Removed fraction
    Remediation Option Chemical 1 Chemical 2 Chemical 3
    ROmax 1 1 1
    RO1 0.739 0.581 0.520
    RO2 0.417 0.240 0.141
    RO3 0.852 0.623 0.909
    RO4 0.035 0.011 0.603
    RO5 0.065 0.689 0.413
    RO0 0 0 0
     | Show Table
    DownLoad: CSV

    Table 1 is the basic dataset for the present study; the single values correspond to the fraction of the 3 chemicals, respectively, being removed by the 5 remediation options.

    To illustrate the effect of specific improvements of the remediation of a single option we assume in a first approach that the RO5 could be improved in a way so Chemical 1 could be 56.5% removed (Table 2).

    Table 2. Fraction of three chemicals being removed by different remediation options improving RO5 to better remove Chemical 1
    Removed fraction
    Remediation Option Chemical 1 Chemical 2 Chemical 3
    ROmax 1 1 1
    RO1 0.739 0.581 0.520
    RO2 0.417 0.240 0.141
    RO3 0.852 0.623 0.909
    RO4 0.035 0.011 0.603
    RO5 0.565 0.689 0.413
    RO0 0 0 0
     | Show Table
    DownLoad: CSV

    To further elucidate the effect of a virtually equal removal of the 3 chemicals we assume an approach where the 3 chemicals are removed virtually to the same extent by RO1 (Table 3).

    Table 3. Fraction of three chemicals being removed by different remediation options following improvement of RO1 to removed Chemical 2 and 3 to an equal extent
    Remedied fraction
    Remediation Option Chemical 1 Chemical 2 Chemical 3
    ROmax 1 1 1
    RO1 0.739 0.781 0.720
    RO2 0.417 0.240 0.141
    RO3 0.852 0.623 0.909
    RO4 0.035 0.011 0.603
    RO5 0.065 0.689 0.413
    RO0 0 0 0
     | Show Table
    DownLoad: CSV

    It should be noted that the above datasets are constructed in order to illustrate in the best possible way the advantageous use of partial order methodology to select the most appropriate remediation technology and as such the present study can be considered to be methodological.

    3.Results and Discussion

    An obvious way of ordering the different remediation options is by using an aggregated indicator that gives an indication on the overall remediation percentage. This may be done by a simple arithmetic average of the single indicator values for each remediation option. Hence, before exploring the partial order methodology it might be useful to answer the question: why partial ordering and not a simple aggregated indicator that immediately would allow an absolute ordering of the various options? The answer can be found by looking at a rather simple example.

    As already mentioned in the introduction the use of aggregated indicator values in order to achieve an absolute ordering of the objects, here ROs, potentially is subject to compensation effects [3] whereby significant information may be lost. Thus, assume that ROx will clean for the 3 chemicals to 0, 100 and 100%, respectively. An aggregated indicator, based on a simple arithmetic mean will be (0 + 1 + 1)/3 = 0.667, corresponding to an overall remediation percentage of 67.6%. Alternatively we could apply a ROy where the removal of the 3 chemicals were 67.6%, thus by the same type of aggregated indicator, would lead to (0.676 + 0.676 + 0.676)/3 = 0.676, in other words to exactly the same overall remediation percentage, 67.6%, despite the fact that the two options obviously lead to significantly different remediation results, a difference, and the background for the difference that in no way can be explained based on the aggregated indicators. A possible introduction of weights for the single chemicals may further muddle the picture.

    3.1.Partial ordering

    Looking at the data given in Table 1 it is immediately clear that the 5 possible remediation options are quite different in their removal of the different chemicals ranging from 1.1% (RO4/Chemical 2) to 90.9% (RO3/Chemical 3). Further with a look at the data in Table 1 it is further clear that it is not evident which of the 5 remediation options that is actually the most optimal. However, this becomes much more obvious when looking at the partial ordering as displaying as a Hasse diagram in Figure 1.

    First of all it is disclosed from Figure 1 as well as from the MIS in Table 1 that none of the ROs equals the optimal situation, as given by ROmax but, on the other hand, they are all better than the starting point, i.e., RO0.

    Scrutinizing Figure 1, it is immediately seen that RO3 > RO4 and RO3 > RO1 > RO2, thus obviously fulfilling equation 1.On the other hand RO5 apparently is incomparable to all the other four ROs. Thus, we have basically no information about RO5 in relation to one or more of the other 4 ROs. It should be noted that the location of an object like RO5 in the Hasse diagram by convention will be placed as high as possible but in principle it could be placed where ever in the diagram within the limits RO0 and ROmax, respectively. Turning to RO4 it is seen from Figure 1 that it is better than RO0 and worse than RO3.However, if RO4 is better or worse than RO1 or RO2 cannot be deducted from the diagram. To obtain further information on the mutual ordering of the different options we turn to average orders (vide infra).

    It is worthwhile to note that an apparent disadvantage of the partial ordering is that the ordering equation (equation 1) obviously may cause incomparabilities even due to very minor differences in indicator values. The software ELECTRE III takes care for such cases by a set of parameters beyond the data matrix and functions describing the degree of user preferences, which basically means that ELECTRE III can be seen as a fuzzy method [11]. A fuzzy concept is built in the partial order methodology [11]. However, this is outside the scope of the present study. Nevertheless, the possible effect of small differences in indicator values are addressed (vide infra).

    3.2.Average orders

    As evident from the discussion above, certain ROs, as RO4 and RO5 may possess a variety of orders without violating the overall ordering displayed in Figure 1. Using RO4 as an example the following orderings are possible ROmax > RO3 > RO4 > RO1 > RO2 > RO0, ROmax > RO3 > RO1 > RO4 > RO2 > RO0 and ROmax > RO3 > RO1 > RO2 > RO4 > RO0, respectively. Thus, the single ROs may in the partially ordered system be assigned a multitude of orders dependent on the actual location and environment in the partial order, as illustrated by the Hasse diagram (Figure 1). Consequently an absolute ordering of the ROs is not possible. However, it is possible to deduct a weak ordering based on average orders, formally obtained as averages of the probabilities for the single possible ranks [20, 21, 22]. In the present case calculations use the exact method based on lattice theory [15, 16, 17].

    In Table 4 the average order, denoted Rkav, for the different ROs is given together with the [0, 1] normalized Rkav as well as the overall fraction being remedied, calculated as the arithmetic means of the indicator values given in Table 1.

    Table 4. Ordering of the original dataset (Table 1)
    Ordering
    Remediation Option Rkav [0, 1]
    normalized Rkav
    Overall fraction remedied
    ROmax 7 1 1
    RO3 5.800 0.800 0.795
    RO1 4.200 0.533 0.613
    RO5 4.000 0.500 0.389
    RO4 3.400 0.400 0.216
    RO2 2.600 0.267 0.266
    RO0 1 0 0
     | Show Table
    DownLoad: CSV

    The discrepancy between the ordering based on the Rkav and the arithmetic mean values is evident when looking at RO4 and RO2.

    Admittedly the differences in the present rather simple study do not appear spectacular. It should in this connection be noted that in the present study the partial ordering as displayed in Figure 1 has very few incomparabilities. As a consequence of that each aggregation, which is order preserving (such as adding or multiplying indicator values) will reproduce the majority of order relations.

    A further point that should be addressed is the potential problem concerning small differences in indicator values. Thus, looking at the data Table 1, looking at RO1 and RO4, it is seen that RO1 > RO4 for Chemicals 1 and 2, whereas RO4 > RO1 for Chemical 3, which leads to the incomparability between these two options (Figure 1). Based on the average orders (Table 4) it is concluded that RO1 > RO4.However, it may be argued that the difference in indicator values for Chemical 3 is so small that it could be neglected which would make RO1 and RO4 comparable and thus intuitively lead to the identical result, i.e., RO1 > RO4.The question of the possible effect of small differences can be mathematically dealt with applying a specialized module of the PyHasse software [19]. Hence, in the present case, i.e., RO1 vs RO4, it is disclosed that if absolute differences in the [0, 1] normalized indicator values, a threshold ˂ 0.09 is accepted as insignificant RO1 = RO4 with regard to Chemical 3 [19].

    3.3.Improving a remediation technique

    The above data and results may constitute an appropriate background for additional work in order to improve one or more of the remediation technologies in question. For illustrative purpose we will in the following focus on the two options RO1 and RO5, respectively. Figure 1 discloses that RO1 is “locked” between RO2 and RO3 whereas RO5 in principle is able to “float” to any position between the two extremes, RO0 and ROmax. It is further noted (Table 1) that RO1 is rather effective in removing Chemical 1 and medium effective for Chemicals 2 and 3.On the other hand it is clear (Table 1) that RO5 is rather poor in removing Chemical 1, medium effective for Chemical 3 and medium to high effective for Chemical 2, respectively. Despite these differences the average orders of the two options appear rather close (Table 4), the [0,1] normalized values being 0.532 and 0.500 for RO1 and RO5, respectively, whereas a significant discrepancy between the respective overall fractions being remedied are noted (Table 4), which, as explained above does not disclose any details about the two remediation options.

    Assuming that RO5 for some reason, e. g., economic considerations, is favored over RO1, we can use the above described tools to elucidate what changes in the fractions being removed affects the average orders and thus possibly the decision process.

    Obviously the very low removal of Chemical 1 constitutes the immediate problems in relation to RO5.Let us then assume that it is possible to adopt some strategies or methods from one of the other ROs that would improve the remediation percentage for Chemical 1 by RO5 from 6.5% to 56.5%, the resulting perturbed MIS is given in Table 2. This MIS lead to the Hasse diagram visualized in Figure 2.

    Figure 2. Hasse diagram of the first perturbed MIS given in Table 2.

    The change in the diagram in Figure 2 compared to that based on the original MIS (Figure 1) is minor but important as RO5 is now comparable to RO2.Thus, the possible orders for RO5 have by this been somewhat limited. However, the two ROs to be compared, i.e., RO1 and RO5, are still incomparable.

    Turning to the average orders (Table 5) it is immediately noted that the change in the remediation percentage for Chemical 1 by RO5 significantly changed the mutual relation between the options RO1 and RO5 the latter now being the more attractive choice.

    Table 5. Ordering of the first perturbed MIS (Table 2)
    Ordering
    Remediation Option Rkav [0, 1]
    normalized Rkav
    Overall fraction remedied
    ROmax 7 1 1
    RO3 5.727 0.788 0.795
    RO5 4.636 0.606 0.556
    RO1 4.000 0.500 0.613
    RO4 3.364 0.394 0.216
    RO2 2.273 0.212 0.266
    RO0 1 0 0
     | Show Table
    DownLoad: CSV

    A final remark on the relation between RO1 and RO5, from Table 2 it can be seen that RO1 > RO5 for Chemical 1 and 3 whereas RO5 > RO1 for Chemical 2.Now it may be argued that the differences for indicator values Chemical 2 and 3 are − 0.108 and + 0.107, respectively and thus they “somehow” compensate for each other and leave only the indicator Chemical 1 as determining which would leave to the opposite ordering conclusion, i.e., RO1 > RO5.However, this is exactly a conclusion that cannot be drawn as there is no proof that what are good/bad for Chemical 2 can be compensated by what are bad/good for Chemical 3 [3].

    An alternative approach would obviously be to improve RO1 by increasing the remediation percentages of Chemicals 2 and 3 by, e.g., 20% thus have a remediation option that for all 3 chemicals would lead to roughly a 75% removal. The corresponding MIS is given in Table 3 which leads to the Hasse diagram displayed in Figure 3.

    Figure 3. Hasse diagram of the second perturbed MIS given in Table 3.

    Comparing to the original Hasse diagram (Figure 1) it is obvious that significant changes have now appeared with an increased number of comparisons. Thus, it is immediately noted that now RO1 and RO5 are comparable and that RO1 > RO5. It can further be noted that now RO1 is no longer comparable to RO3 and both RO1 > RO4 and RO3 > RO4.

    The fact that RO1 > RO5 is accordingly also shown when looking at the average orders (Table 6), where RO1 now constitutes as the most optimal remediation option. Hence, assuming all other equal RO1 should be chosen for the given remediation task.

    Table 6. Ordering of the second perturbed MIS (Table 3)
    Ordering
    Remediation Option Rkav [0, 1]
    normalized Rkav
    Overall fraction remedied
    ROmax 7 1 1
    RO1 5.571 0.762 0.747
    RO3 5.286 0.714 0.795
    RO5 3.286 0.381 0.389
    RO2 2.929 0.322 0.266
    RO4 2.929 0.322 0.216
    RO0 1 0 0
     | Show Table
    DownLoad: CSV

    It is at this stage necessary to comment on the term “all other equal”. In the present study we have considered the removal of 3 chemicals. The original as well as the residual “concentration” of the chemicals may be given as amounts (e. g., g/kg), toxicity (e. g., LD50 towards a selected organism), the carcinogenic potential etc. However, it is obvious that other factors typically are brought into play such as economic aspects, environmental aspects as destruction of areas worthy of preservation, destruction of habitats etc.

    It is immediately possible to apply the above described methodology on these aspects as well and eventually combine the analyses applying hierarchical partial ordering [23] However, this is outside the scope of the present study.

    4.Conclusion

    In the present study a simple analytical tool for decision support in relation to selecting the more appropriate remediation technology/option for polluted sites. It has been demonstrated that applying partial order methodology a deeper insight in the factors governing the different options can be disclosed and as such form the background, not only for selecting the more appropriate option but also serving as an information source in relation to further improvements of one or more of the options under discussion.

    The here presented example is based on the remediation of 3 chemicals. Obviously these chemicals, and thus the values in the MIS can be any kind of measure of “concentration”, i.e., amounts, persistence, bioaccumulation, toxicity etc. As such presented method does not have any imitations. Further, factors like economic and environmental considerations may be taken into account, possibly in a separate approach eventually combining the various analyses through a hierarchical approach [23].

    Finally it should be emphasized that the ratio between the number of objects studied and the number of indicators may turn out as crucial [20]. Thus, if the number of indicators relative to the number of objects studied is too high the number of incomparisons unambiguously will increase, ultimately leading to a complete antichain obviously with a loss of predicting power. However, this apparent problem may be circumvented by applying hierarchical partial order ranking (HPOR) as previously described by the author [23]. Here the group of indicators is subdivided into appropriate groups. The objects, here remediation options are thus partially ordered according to the single subgroups of indicators. In a second partial ordering the results of the original analyses of the single subgroups are combined to the final ordering.

    Conflict of Interest

    Author declares no conflicts of interest in this paper.

    [1] Math. Med. Biol., 25 (2008), 185-186.
    [2] Nature Reviews Cancer, 8 (2008), 227-244.
    [3] Parasitology, 85 (1982), 411-426.
    [4] Science, 303 (2004), 842-844.
    [5] Adv. Cancer Res., 76 (1999), 187-212.
    [6] Phi. Trans. R. Soc. A, 364 (2006), 1563-1578.
    [7] CRC Press, 32, 2009.
    [8] Acta Biotheoretica, 58 (2010), 329-340.
    [9] CRC Press, 34, 2010.
    [10] in press.
    [11] Birkhauser, 2005.
    [12] J. Stat. Phys., 128 (2007), 287-345.
    [13] Parasitol Today, 12 (1996), 96-101.
    [14] Evolution, 51 (1997), 1828-1837.
    [15] Milan: Springer, (2006), 71-108.
    [16] Q. Rev. Biol., 71 (1996), 37-78.
    [17] Cytometry, 69A (2006), 704-710.
    [18] Nature, 421 (2003), 321.
    [19] CELL, 100 (2000), {57-70}.
    [20] Mathematical Biosciences and Engineering: MBE, 6 (2009), 521.
    [21] Genetics, 166 (2004), 1571-1579.
    [22] Bulletin of the American Physical Society, 56 (2011).
    [23] Jour. Stat. Phys., 128 (2007), 413-446.
    [24] Bull. Math. Biol., 68 (2006), 1573-1599.
    [25] J. Theor. Biol., 223 (2003), 433-450.
    [26] Emerg. Infect. Dis., 2 (1996), 93-102.
    [27] Nat. Rev. Cancer, 6 (2006), 924-935.
    [28] Cell Cycle, 3 (2004), 358-362.
    [29] Clarendon, Oxford, 1962.
    [30] Science, 303 (2004), 793-799.
    [31] Proc. Natl. Acad. Sci. U S A, 99 (2002), 16226-16231.
    [32] Proc. Biol. Sci., 255 (1994), 81-89.
    [33] Proc. Natl. Acad. Sci. U S A, 101 (2004), 10635-10638.
    [34] Science, 194 (1976), 23-28.
    [35] Sem. Cancer Biol., in press, (2008).
    [36] Wiley Interdisciplinary Reviews: Systems Biology and Medicine, 3 ( 2011), 115-125.
    [37] CLINICAL CANCER RESEARCH, 12 (2006), 3882-3889.
    [38] Multiscale Cancer Modeling, 34 (2010), 173.
    [39] Ecology Letters, 6 (2003), 176.
    [40] Cambridge University Press, 1982.
    [41] Biology Direct, 5 (2010), 21.
    [42] Cambridge University Press, 2005.
    [43] Int. J. Epidemiol, 35 (2006), 1151-1159.
    [44] World Scientific, 2005.
    [45] in "Proceedings of the Sixth International Congress on Genetics", (1932), 355-366.
    [46] Journal of Mathematical Biology, 58 (2009), 545-559.
  • This article has been cited by:

    1. Tao Mi, Xiangpan Kong, Meiling Chen, Peng Guo, Dawei He, Inducing disulfidptosis in tumors:potential pathways and significance, 2024, 5, 2688-2663, 10.1002/mco2.791
    2. Dong Guo, Zhihao Liu, Jinglin Zhou, Chongrong Ke, Daliang Li, Significance of Programmed Cell Death Pathways in Neurodegenerative Diseases, 2024, 25, 1422-0067, 9947, 10.3390/ijms25189947
    3. Shanpeng Liu, Qike Wu, Can Xu, Liping Wang, Jialing Wang, Cuiying Liu, Heng Zhao, Ischemic Postconditioning Regulates New Cell Death Mechanisms in Stroke: Disulfidptosis, 2024, 14, 2218-273X, 1390, 10.3390/biom14111390
    4. Arian Daneshpour, Zoha Shaka, Nima Rezaei, Interplay of cell death pathways and immune responses in ischemic stroke: insights into novel biomarkers, 2024, 0334-1763, 10.1515/revneuro-2024-0128
    5. Sijun Li, Ningyuan Chen, Junrui He, Xibao Luo, Wei Lin, NDUFA11 may be the disulfidptosis-related biomarker of ischemic stroke based on integrated bioinformatics, clinical samples, and experimental analyses, 2025, 18, 1662-453X, 10.3389/fnins.2024.1505493
    6. Shuang Wang, Xinhua Liu, Jun Tian, Sizhu Liu, Lianwei Ke, Shuling Zhang, Hongying He, Chaojiang Shang, Jichun Yang, Zhiwen Luo, Bioinformatics analysis of genes associated with disulfidptosis in spinal cord injury, 2025, 20, 1932-6203, e0318016, 10.1371/journal.pone.0318016
    7. Jing Chang, Danhong Liu, Yuqi Xiao, Boyao Tan, Jun Deng, Zhigang Mei, Jun Liao, Disulfidptosis: a new target for central nervous system disease therapy, 2025, 19, 1662-453X, 10.3389/fnins.2025.1514253
  • Reader Comments
  • © 2013 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2917) PDF downloads(608) Cited by(8)

Article outline

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog