Three words sum up the requirements for the cure of cancer, comprehensiveness, specificity and knowability. All three requirements must be jointly satisfied. Comprehensiveness is about killing all cancer cells that could evolve. Specificity is about doing so without producing unacceptable side effects. Knowability refers to the need to target properties that are known or that can be known, with the logically sound formulation of scientific knowledge as defined by Popper.

Replace the word killing with controlling and you have the requirements for the chronic control of cancer. The requirements for cure and control are virtually the same.

A thought experiment

Let us imagine a machine that can specifically and consistently cure metastatic cancer. How must any such cancer-curing machine work? What are the logical processes and requirements that must be satisfied for such a hypothetical machine to perform its task? Our goal is to discover principles of nature that apply to all possible methods for the specific cure of cancer. With an understanding of such principles we can then develop a rational scientific strategy to specifically cure cancer. We place only one constraint our hypothetical machine: it must obey known fundamental laws of nature.

A hypothetical cancer-curing machine

In the tradition of the best scientists let’s use our imagination and consider a hypothetical “cancer cell detection and destruction machine” that can consistently and specifically cure cancer. In the spirit of the best reductionist traditions of science, let us take our machine apart and see how it works.

The machine functions by: Asking questions about the cells and their environments Receiving bits of information in the form of answers [i] Deciding: normal cell or cancer cell. (spare or kill) Killing only the cancer cells

If you were writing a computer program to control the cancer-curing machine these are the essential steps that would be required. However, let us examine them in more detail. We are interested in creating a theory of the requirements for the cure of cancer that rests not upon the obvious, but upon deductive logic and well-established principles of nature. Steps 1 and 2 follow from Shannon’s Theory of Information and fundamental laws of physics. Steps 3 and 4, deciding cancer or not, and killing only the cancer cells, follow purely by deductive logic from the requirement of tumor specificity.

It follows from Shannon’s Theory of Information and fundamental laws of physics that our hypothetical cancer-curing machine requires bits of information to function. [1]

A primer on information

One bit is the information derived from a binary (yes or no) question in which the expected outcomes are equally probable to the receiver of the information. If I flip a coin, and tell you the outcome is heads, then you will have just received one bit of information. That is the amount of information. If we agreed beforehand that heads means you win, and the result is heads, then the message conveyed or the content of the information is, you win. There is a difference between the quantity of information and the content of the information. The quantity of information has an objective physical reality. It is as real as a calorie or watt-hour of energy. Most people are aware that it takes a certain number of bits of information to store a computer document and that CDs can hold a certain number of bits of data. You cannot see a bit of information, it’s an abstraction, and yet information has an underlying objective reality.

The fundamental role of information in the cure of cancer

Let us now return to our hypothetical cancer-curing machine. The logical requirements for the specific cure of cancer are principally about the information required for the comprehensive and specific detection and destruction of all malignant cells that could revolve in the patient. An analysis of the questions that must be asked, and the information that must be received, for the machine to perform its task defines the fundamental requirements for the cure of cancer. The minimum information requirements for this hypothetical machine to detect any malignant cell that could evolve, are the essential information requirements for the consistent and specific cure or control of cancer. Any therapy that can consistently and specifically cure or control metastatic cancer must function in a manner that is logically equivalent to this hypothetical “cancer cell detection and destruction machine” and requires the same or logically equivalent information. It does not matter if the therapy is based on drugs, the immune system , viruses, monoclonal antibodies, “smart bombs,” nanopaticles, or radiation. The information requirements that must be satisfied are the same.

Logically equivalent information

The concept of logically equivalent information needs some explanation. By this we mean that the information content conveys the same message. For example, consider a jar with two kinds of marbles some blue and some hybrid marbles each of which is half red and half yellow. To sort these marbles you could examine each marble and ask one of four questions:

• Is the marble blue?

• Is the marble half yellow?

• Is the marble half red?

• Is the marble both yellow and red?

The answer to any of one these questions about a particular marble allows all of the questions to be answered. The same holds true for our hypothetical cancer-curing machine. There are very many different ways in which the questions can be asked, but in the end the same amount of information is required and the information must be logically equivalent. It must convey the same message, cancer or normal, spare or kill.

The information required to sort or distinguish any malignant cell from any normal cells is of fundamental importance to the specific cure of cancer. How many bits of information are required? What questions must be asked by our hypothetical cancer-curing machine?

Binary questions

For the sake of simplicity let us confine our machine to asking questions that can be answered as either yes or no (0 or 1). In other word our machine asks binary questions and operates with digital information. This poses no loss of generality. Modern computers deal in binary digital information and can solve an almost unlimited range of problems. The mathematician Alan Turing proved that any computationally solvable problem can be expressed in binary digital information. So we are on very solid ground by restricting our hypothetical cancer-curing machine to binary questions.

Hidden assumptions in the operation of our hypothetical cancer-curing machine

There are several hidden assumptions embedded within the logical operations of our machine, which are repeated below:

1. Asking questions about the cells and their environments

2. Receiving bits of information in the form of answers

3. Deciding: normal or cancer (spare or kill)

4. Killing only the cancer cells

Step 1 contains the assumption that the machine knows what questions to ask. Step 3 implies that the machine knows a rule for using the information to decide: cancer or not, spare or kill. Step 4 implies that the machine knows how to reliably kill any cell deemed to be a cancer cell. For our machine to function, these issues must be addressed.

Quantitative information requirements for a cancer-curing machine

It is instructive to begin by asking the question, How much information is needed for our cancer-curing machine to function? When our machine asks a question, there are just two possible answers, yes or no. Both are equally likely. [ii] Accordingly, as in the case of a coin flip, the answer provides one bit of information.

One bit or multiple, logically-unconnected, single bits of information cannot work There is only one class of questions that could potentially enable the machine to perform its task on the basis of one bit of information. The only question that can specifically identify cancer cells using one bit of information is the logical equivalent of: Is the cell a cancer cell? This is equivalent in the language of molecular biology and chemistry to asking: Does the cell have a known tumor-specific molecule or tumor-specific molecular lesion (such as Bcr-Abl)?

To ask this question requires specific prior knowledge related to a pathway of tumor cell evolution. For example, the machine must know about Bcr-Abl in order to ask the question, is Bcr-Abl present? The machine must know how to detect Bcr-Abl.

Let’s assume our machine can ask the question, is Bcr-Abl present. If the answer is yes then the machine will decide the cell is a cancer cell and kill it. However, if the answer is no then our machine cannot conclude that the cell is normal. The cell could be a cancer cell that evolved without Bcr-Abl. Or the cell could have a mutated form of Bcr-Abl that is not detectable.

Now you might say, how about if our machine asks another question about a different tumor specific lesion. It can do that but we still have the same problem. If the answer is no, the machine still remains uncertain and could allow a cancer cell to escape destruction. To perform its task of specifically curing cancer our machine would have to ask a question about every pathway of tumor cell evolution that could realistically occur in the patient. To do so, the machine must have prior knowledge of these pathways. However, as previously discussed the number of evolutionary pathways is almost unlimited. Comprehensive knowledge of the pathways of tumor cell evolution in a patient is generally unknowable. In other words, our machine cannot know what questions to ask.

It logically follows, that no machine or process can consistently and specifically cure cancer by asking one-bit questions. Nor can our machine chronically control cancer by asking one-bit questions. It will fail.

The failure rate

The failure rate will depend upon the number of cancer cells in the patient and the degree of genetic and epi-genetic heterogeneity. A pea-sized lesion has about one billion cancer cells. Consider a tumor specific molecular target that is present in 99.999% of the malignant cells in a patient. [iii] This would mean that only one out of 100,000 cells lacked the target. Our machine would fail to detect and kill 10,000 cancer cells in such a pea-sized tumor. In theory, one cancer cell that escapes detection and destruction could expand to give one trillion cells after only forty doublings and result in fatal disease.

One-bit questions can provide specificity but cannot provide comprehensiveness

Our thought experiment so far has taught us a very important principle about cancer therapy. The principle is a direct logical consequence of the astronomically diverse, stochastic, evolutionary nature of cancer. It applies to all possible approaches to the consistent and specific cure or control of the disease. It applies to all possible types of anticancer drugs and therapies.

Any cancer therapy in which tumor cell identification is based on individual molecular targets will fail. Any drug or cancer therapy that identifies cancer cell by asking one question and receiving one bit of information will fail. Any reasonably finite set of drugs or therapies that identify cancer cells on the basis of single bits of information will fail.

Anti-cancer drugs can be made that function by asking a one-bit question

For example, the drug Iressa, which inhibits certain mutant EGF receptors found on some cancer cells, can be regarded as asking the question, Is the mutant EGF receptor present, and then performing the logical operation, if yes, then attack the cell; if no then spare the cell. Such one-bit drugs can shrink some tumors or slow cancer progression. In some patients the results, although temporary, can be dramatic. However, such drugs cannot with any degree of consistency cure or chronically control cancer.

Objection

The drugs Gleevec targets the tumor specific molecule, Bcr-Abl and is quite effective in controlling chronic myelogenous leukemia (CML). Your whole point is wrong.

Response

Not all. Resistance develops to these drugs. In addition, CML at least initially has a very slow rate of evolution. Most cancers present with far greater genetic diversity and have much greater rates of tumor cell evolution. It is a grave error to conclude that drugs similar to Gleevec can work for metastatic cancer in general. Such a conclusion is based on the false logic of induction and is logically inconsistent with tumor cell evolution.

Multiple-bit questions can provide both specificity and comprehensiveness Let us examine more closely the requirements for our cancer-curing machine to function. The cancer-curing machine does not need to kill all tumor cells. Only malignant cells can sustain cancer. Therefore the detection and destruction of all cells that engage in malignant behavior is sufficient to cure cancer. As previously discussed, all malignant cells use normal cellular machinery to carry out malignant behavior. The normal cellular machinery that can carry out malignant behavior can be known and largely already is known. Furthermore, this is the only class of information that can be known that is comprehensive and specific to all malignant cells. This implies that our cancer-curing machine must function by asking at least two logically connected questions that together detect proliferation and invasiveness on the basis of normal cellular machinery. Each of the questions that together form a compound question requires one bit of information to answer. An example follows: Does the cell and/or its environment have a pre-defined, normal protein that is characteristic of cell proliferation, and another pre-defined, normal protein that is characteristic of invasiveness ? [iv] The answer to this multi-bit question requires two bits of information. If the answer is yes to both parts of the question, then the machine will consider the cell malignant and kill it. Otherwise the cell is spared.

We will get to the all-important issues of specificity and comprehensiveness. However, to provide an overview we will first briefly touch upon a few crucial points.

The detection of malignant behavior

Malignant cells cannot be identified solely on the basis of proliferation. Proliferation is a normal cellular function vital to life and a characteristic of many normal cells. Similarly, invasiveness alone is a normal cellular activity. However, both proliferation and invasiveness are very tightly regulated physiological processes. The combination of proliferation and invasiveness is highly restricted [v] and occurs in “normal” non-malignant settings such as wound healing, menstruation, placental implantation, tissue remodeling, new blood vessel formation, embryonic and fetal development, and infectious processes such as abscess formation. In abnormal settings, proliferation and invasive define malignant behavior.

It is also important to emphasize that invasiveness is not a property only of the malignant cell. It is also a property of the cancer cell’s environment and of time. The normal cellular machinery that carries out invasiveness can be made both by cancer cells and by normal cells recruited to the tumor by the cancer cells. It is for this reason that the word "environment" is included in the prior multi-bit question. Just as it is not meaningful to speak of a single molecule of water as boiling, it is not meaningful to speak of a cell in isolation as expressing invasiveness.

If our hypothetical machine is turned off during times of physiological invasiveness , such as pregnancy and wound healing, then the detection of the combination of proliferation and invasiveness can enable the specific detection of malignant behavior and malignant cells.

In many cases the detection of invasiveness alone would suffice. Invasiveness will be discussed in detail later. It should be noted that the detection of invasiveness generally requires asking a multi-bit question. No single type of protein is currently known that is absolutely characteristic of invasiveness.

Let us now return to our prior multi-bit question:

Does the cell and/or its environment have a pre-defined, normal protein that is characteristic of cell proliferation, and another pre-defined, normal protein that is characteristic of invasiveness ?

The term “pre-defined” means that the machine knows or has been programmed with the information required to ask the questions. The machine is looking for something that is non-random, well characterized and known beforehand. This is doable because the machine is focusing on normal proteins and normal cellular machinery. It is not dealing with the chaos of cancer. It is focused on the boundary conditions, the canyon walls that constrain the flow of tumor cell evolution.

Detecting a pattern of proteins  The process of asking the prior multiple-bit question is logically equivalent to detecting the presence or absence of the corresponding combination or pattern of normal proteins. In other words we could rephrase the previous question as follows: Does the cell and/or its environment have a pre-defined, pattern of normal proteins that is characteristic of cell proliferation and invasiveness?

The term “pattern” is used to refer to the set or combination of proteins, not to any particular spatial arrangement.

The concept of detecting a pattern is simple. The pattern is present if and only if all members of the pattern are present. [vi] If we consider the different kinds of proteins as letters of the alphabet, then a pattern of proteins  is a word. Just as words, not letters, have meaning, patterns, not individual proteins, are the target.

 

 

Tumor specificity can reside in a pattern, but not in the individual normal proteins that comprise the pattern. Normal proteins are normal and not tumor specific. The information is in the pattern, not the proteins that comprise the pattern.

Targeting a pattern: Pattern Recognition Tumor Targeting

Targeting specificity is for a pattern, not the individual proteins that comprise the pattern. [vii] For example, a toxic drug targeted to the pattern of the two proteins A and B would only kill cells that express both proteins. Cells that express only one of the proteins would be spared. We call approaches that target patterns Pattern Recognition Tumor Targeting (PRTT).

PRTT is multi-variable targeting

Another way to explain the concept of PRTT is to consider two binary mathematical functions: F(x) = 0 or 1 and G(x,y,z) = 0 or 1. The variables x, y, and z relate to three different cellular properties. The value 0 means normal cell, spare. The value 1 means cancer cell, kill.

The function F(x) has one variable and the output depends only on the value of x. F(x) does not “target” a pattern. By contrast, G(x,y,z) has three variables that determine the output and “targets” the pattern (x,y,z). PRTT is about making drugs that operate like the function G(x,y,z). (In general, the number of variables or elements in the target patterns of PRTT-based drugs will be 2 or 3.)

Degeneracy

Not all patterns of proteins that are “characteristic” of proliferation and invasiveness are specific for malignant behavior. For example, the protein, urokinase, is involved in invasiveness and expressed by many types of malignant cells. [2] However, normal kidney cells, which are not invasive, also express urokinase. [3] The reason for this is that biological systems are degenerate.

Dr. Gerald Edelman has written extensively about biological degeneracy and its importance in evolution. [4] There are often multiple, totally different proteins that can perform the same function. And a single protein can also have multiple, totally different roles depending upon the context. This is degeneracy. An engineer designs a structure to serve a particular function. Evolution doesn’t work that way. In the words of Dr. Edelman:

In engineering systems, logic prevails, and, for fail-safe operation, redundancy is built into design. This is not the case for biological systems. Indeed, not the least of Darwin's achievements was to lay the argument by design to rest. In general, an engineer assumes that interacting components should be as simple as possible, that there are no “unnecessary” or unplanned interactions, that there is an explicit assignment of function or causal efficacy to each part of a working mechanism…Irrelevancy is avoided from the outset. By contrast, in evolutionary systems, where there is no design, the term “irrelevant” has no a priori meaning. It is possible for any change in a part to contribute to overall function, mutations can prompt compensation, stochastic interactions with the environment can lead to strong selection, often there is no fixed assignment of exclusive responsibility for a given function, and, unlike the engineering case, interactions become increasingly complex… [4]

Consider a car. Different parts have distinct well-defined functions. The fuel pump pumps the fuel, the spark plugs ignite the fuel and the wheels rotate and allow the car to roll. To an extent, the same also holds true for biological systems. The heart pumps blood, the blood vessels carry blood, the lung oxygenates the blood. A clear assignment between structure and function is possible, most of the time, but not always. Evolutionary changes occur in the context of an entire system of interacting parts. As a result the same protein can often have multiple unrelated functions. Proteins that are involved in vital, indispensable cellular functions often can be lost without adverse consequence. Other proteins can compensate. This is what degeneracy  is about.

Degeneracy has important implications for the operation of our cancer-curing machine. It implies that the questions asked to detect proliferation and invasiveness must relate to the most important, most basic, most downstream proteins that actually carry out or reflect the processes of malignant behavior.

Let us return to our car analogy. A car has thousands of components. However, to determine if a car is in operation you need only look at a few critical parts such as: the pistons; drive shaft; or wheels. In the same sense to tell if a cell is engaged in malignant behavior one need only look at the critical machinery that actually carries out the final process.

Proteins that have been evolutionarily conserved for long periods of time are the best. Since biological systems are degenerate, care must be taken to exclude patterns of proteins that are present in normal tissues. Even though normal tissues do not engage in malignant behavior, certain proteins patterns “characteristic of proliferation and invasiveness ” may appear in normal tissues. Since biological systems are degenerate, we should expect this to occur.

The key is to exclude patterns that are present in normal tissues.

This is doable and is the goal of the Project for Target Pattern Selection. This information is knowable and can be readily obtained using routine technology to examine pattern expression in normal tissues. The information is not dependent upon the chaos of cancer since it involves well-defined normal proteins and tissues.

We should now update our diagram about what is unique and knowable about cancer.

Specificity and Comprehensiveness

Let us now return to the critical issue of comprehensiveness for our cancer-curing machine.

If the pattern of proteins “XY” is absent from normal tissues then the pattern XY is tumor specific. However, genetic alterations in cancer cells could cause one or both proteins to be lost in any given cell. Therefore, the pattern XY cannot provide comprehensiveness.

Any protein can be lost by mutation

Since cancer is an evolutionary process, for any given cell, any protein can be lost due to genetic or epi-genetic alterations. Even vital evolutionary conserved proteins that are essential for cell survival can be lost to detection. The protein may remain, but cancer cells can evolve that enable the protein to evade detection. In addition, we cannot know what proteins and target patterns any given cancer cell in a patient will lose. This would require comprehensive knowledge of the pathways of tumor cell evolution, which is unknowable. However, we don’t need to know that.

The need for a comprehensive set of target patterns

What we need to know is a set of different target patterns that is sufficiently large so that the probability a cancer cell will evolve without at least one of the patterns is clinically insignificant. [viii] This we can do. Comprehensiveness requires asking the multiple-bit question in a sufficient number of versions, such that the probability that a malignant cell could evolve that evades detection by all versions, is clinically insignificant (i.e., less than 10^-15 per cell division).

As discussed previously, we cannot know the relevant probabilities. However, we don’t need to know them. At the extreme limits of improbability there is near certainty of knowledge. The trick is to employ a sufficiently large number that we are confident that loss of all the target patterns will not occur. Based on known maximal rates of mutation and knowledge, albeit incomplete, of the relevant normal cellular machinery, we estimate that approximately 5 to 10 different patterns will be required for comprehensiveness. The exact number must be experimentally determined.

A comprehensive set of target patterns can be known

The important point is that a set of multiple-bit questions can enable comprehensive cancer cell detection based on patterns of normal cellular machinery. [xxi] These patterns can be known beforehand and are independent of the astronomically diverse stochastic pathways of tumor cell evolution. These patterns are common to all forms of solid cancers.

By contrast, it is not possible to pre-define a comprehensive set of tumor targets that is based on stochastic pathways of tumor cell evolution.

From a logical point of view there is little difference between a known, mutant, tumor-specific protein and a known, tumor-specific pattern of normal proteins. Once a particular genetic alteration is known, it is in a sense, no longer random. The major distinction is that we can know a set of patterns that will be comprehensive for all malignant cells that could evolve. We can know this by deductive logic, and by empirical observation of the normal cellular machinery of proliferation and invasiveness. By contrast, it is not possible to know a set of tumor specific genetic alterations that will be comprehensive for all malignant cells that could evolve in the patient. It is difficult to over-emphasize the importance of this point.

There is a concrete difference between normal cells and malignant cells.

Normal cells do not engage in malignant behavior, by definition all malignant cells do. This concrete difference must be reflected by normal cellular machinery actually in the process of carrying out or executing malignant behavior. It does not matter what causes cancer. The genetic and epi-genetic alterations are irrelevant. What matters is what’s knowable. The patterns of normal cellular machinery that carry out malignant behavior are knowable and specific to malignant cells. [ix] [x] [xi]

Malignant behavior can only be detected on the basis of patterns of biomolecules

No single molecular entity or single physical property can enable the detection of malignant cells or malignant behavior. Cancer is not like temperature. This is an extremely important point. Malignant behavior, which is defined as proliferation and invasiveness in an abnormal context, can only be detected on the basis of patterns of biomolecules related to the cancer cell and its environment. The same holds for the detection of malignant cells. [xii]

Multiple patterns are required for comprehensiveness We are now in an improved position to define the requirements for our hypothetical cancer-curing machine to perform its task, the specific cure of cancer. Our machine must ask a series of questions of the following form: [xiii] Does the cell and or the environment of the cell have the abnormal Pattern A of biomolecules that effects or reflects proliferation and invasiveness ;or Does the cell and or the environment of the cell have the abnormal Pattern B of biomolecules that effects or reflects proliferation and invasiveness ; or Does the cell and or the environment of the cell have the abnormal Pattern C of biomolecules that effects or reflects proliferation and invasiveness ...or Does the cell and or the environment of the cell have the abnormal Pattern N of biomolecules that effects or reflects proliferation and invasiveness?

If the answer to any of these questions is yes than the machine decides the cell is a cancer cell and kills it. The number, “n” of patterns must be sufficiently large so that it is too improbable for a cancer cell to evolve without at least one of the patterns being present and detectable.

In practical terms, to satisfy these requirements a drug will be needed that can detect and destroy cells that express each pattern. The “n” drugs must be given in combination.

The set of patterns must be comprehensive.

The set of patterns (A,B,C…N) must be selected so as to provide comprehensiveness. [xiv] In other words, the patterns must be selected such that any malignant cell (or its environment) must express at least one of the patterns, at some point in time.

Comprehensiveness in a set of target patterns

Cells often have multiple ways or pathways by which certain activities can be carried out. For example, there are a large number of different types of cell surface receptors and growth factors that can trigger cells to divide. The signals to divide can be transmitted to the DNA of cells by a large number of different routes or biochemical pathways. [xv]

However, there is a convergence. All roads lead to the same place. Ultimately, the actual process of DNA replication and cell division is carried out by evolutionarily conserved, normal cellular machinery that is absolutely required for cell replication. This machinery is common to, and characteristic of, all pathways of cell replication. Accordingly, the detection of a small number of proteins is sufficient to comprehensively identify any cell that is replicating. In other words, to know with certainty that a cell is replicating, our cancer-curing machine needs to ask only a small number of questions. Based on known mutation rates, two to three different questions should be sufficient for the comprehensive detection of proliferation.

The comprehensive detection of invasiveness is a bit more involved. Not all pathways of invasiveness converge. For example, many different types of enzymes can degrade collagen and promote tumor cell invasion. Proteins or markers are needed that can detect each independent pathway of invasiveness. However, most pathways are not truly independent. There are good reasons to expect that a relatively small number of protein patterns will be sufficient to comprehensively detect invasiveness. (The comprehensive inhibition of invasiveness is another story and would require targeting a large number of proteins.)

The same set of patterns for all solid cancers

The patterns A, B,C… N can and must be experimentally defined beforehand. The same exact patterns will apply for all solid cancers. There is no need to individualize the patterns to the particular tumor type or particular patient. [xvi] This follows from the point made in chapter 7 about cancer being essentially one disease. To re-iterate, all solid cancers can potentially use the same normal cellular machinery to carry out malignant behavior. It follows that to achieve comprehensiveness the patterns that must be targeted are identical for all solid cancers.

Redundancy in mechanisms of cancer cell killing 

The information requirements for our cancer-curing machine relate largely to the process of malignant cell identification. Cancer cells can evolve that are able to evade killing by virtually any single drug or toxin. [xvii] A sufficient number of independent methods of cell killing must be simultaneously employed so that it is just too improbable for a cancer cell to escape death. This poses no real problem. In practical terms, about three different toxic agents will be required to achieve comprehensiveness in cell killing.

The purpose of the toxic agents is to kill cells that have been identified as malignant. The toxic agents do not provide the specificity needed to identify the cancer cells.

The selection of a set of toxic agents that is comprehensive poses no special technical challenges. It is well within the scope of existing science and technology. The same set will work for all cancers.

The importance of time

Our hypothetical cancer-curing machine must examine every cell in the body, and it must do so repeatedly for a prolonged period of time. [xviii] The reason for this is that malignant behavior is expressed episodically by cancer cells. The component functions of malignant behavior, proliferation and invasiveness, are also episodic and generally not expressed at the same time. Most malignant cells are not actively proliferating most of the time. Similarly, most malignant cells are not actively engaged in invasiveness most of the time. However, at some point in time the cells must express these features or by definition the cells are not malignant. A prolonged period of time is therefore critical to the function of our machine.

Although malignant cells need not engage in the biochemistry of proliferation and invasiveness at the same time, the combination of invasiveness and the potential for proliferation are expressed simultaneously.

Most normal cells lack the potential to proliferate. MCM proteins are vital, evolutionarily-conserved proteins that serve as excellent markers for the potential to proliferate. Accordingly, invasiveness and the potential for cell proliferation are more efficient markers for the detection of malignant cells.

In practical terms, the specific cure of cancer will require the systemic (intravenous) administration of a set of drugs targeted to the patterns for a prolonged period of time (months).

The required rate of cancer cell destruction

Our cancer-curing machine need not operate continuously. But it must operate for enough time and rapidly enough to kill malignant cells faster than the malignant cells proliferate. Otherwise, the cancer will progress despite the machine. As discussed in another section a minor but sustained decrease in the probability of cancer cell survival can have a truly enormous impact.

The requirements for the chronic control of cancer The requirements for our hypothetical cancer-curing machine are the requirements for the specific cure of cancer. The requirements for the chronic control of cancer are identical with one minor difference. Instead of killing the malignant cells, a “cancer-controlling machine” keeps the cells in check and controls the number of malignant cells below a certain disease-causing threshold. The information requirements are essentially identical for both the specific cure and the specific chronic control of cancer.

Different sides of the same coin

You may be wondering about Dr. Folkman’s pioneering and brilliant work on angiogenesis and starving tumors by cutting off their blood supply. Can’t this chronically control cancer? In theory it can. However, the information requirements to do so are exactly identical to those described above for our hypothetical cancer-curing machine. Inhibition of angiogenesis, or new blood vessel formation, is not sufficient to consistently control cancer.

Two other processes must also be inhibited, vasculogeneic mimicry and the co-option of normal existing blood vessels by tumor cells to consistently starve tumors. Vascular co-option is the invasion of malignant cells along blood vessels. All three of these processes involve the use of normal cellular machinery to carry out proliferation and invasiveness. We are just talking about different sides of the same coin.

The requirements must be satisfied to consistently cure cancer

It should be noted that any process that fails to satisfy the requirements of our hypothetical machine cannot consistently and specifically cure or control cancer. A machine that asks an insufficient number of questions and searches for an insufficient number of target patterns will fail. Any such machine will act a selective pressure and merely redirect the flow of tumor cell evolution. Resistant cancer cells will evolve. The machine will fail. To cure cancer the machine must search for a comprehensive set of target patterns. It must search for cells that express any of the target patterns. It must search for all the target patterns at the same time. Otherwise resistant cells can evolve and the machine will fail.

The immune system cannot target the necessary patterns for cure

Despite billions of dollars and decades of research, attempts to cure cancer with the immune system have met with little success. There are multiple reasons, but a fundamental problem is that the immune system is not able to carry out the operations of our hypothetical cancer-curing machine. The immune system did not evolve with the ability to target patterns of normal proteins that effect or reflect proliferation and invasiveness. The best the immune system can do is to chase after new tumor antigens as they arise. Meanwhile cancer cells can and do evolve mechanisms of escaping immune attack.

Our hypothetical cancer-curing machine addresses the astronomically diverse, unpredictable, stochastic, evolutionary nature of cancer

In principle, the machine can detect and kill all malignant cells that could realistically evolve in the patient while sparing normal cells. Its logical operation is consistent with known fundamental principles of nature. It identifies and destroys cancer cells on the basis of knowable and known information. Its function is independent of any particular pathway of tumor cell evolution, genetic or epi-genetic alteration and genetically encoded molecular target. Our hypothetical cancer-curing machine would work for all forms of solid cancers. [ixx]

The requirements for our hypothetical cancer-curing machine follow by deductive logic from fundamental, well-established, scientific theories and laws of nature. The requirements are very sharply defined because so little can be known about cancer in a patient with metastatic disease. This is a direct consequence of the astronomically diverse, stochastic, evolutionary nature of cancer. The severe constraints to knowledge about cancer leave us with no option but to target all malignant cells that could evolve in the patient. This can only be accomplished by processes that satisfy the requirements of our hypothetical cancer-curing machine.

We freely admit these requirements may be wrong

As Karl Popper emphasizes, “All scientific knowledge is conjecture.” However, to reject these requirements for the specific cure and control of cancer, we must have either logical or empirical reasons that falsify these requirements.

The theory that cancer cells use normal cellular machinery to carry out malignant behavior rests on very solid ground. A violation has never been observed and is far too improbable ever to occur. The theory that all malignant cells can be specifically detected on the basis of abnormal patterns of normal cellular machinery that effect or reflect malignant behavior is also on solid theoretical ground.

Objection

Surely malignant cells can evolve that are very similar to normal cells in which the only chemical differences are far too minor and subtle to detect and target. Maybe just an increase of 1% of a key protein is sufficient to make a cell malignant. Isn’t cancer a continuum? Maybe the only difference will be that the cells have a slightly decreased probability of cell death. After all, minor increases in cell survival can translate into huge growth advantages.

Response

We agree, but only partially. The slope is slippery. There are not always discrete and clear cut transitions from normal to mutant to malignant cell. At times cancers can be low grade. It can be unclear, or even unknowable, if the malignant disease is present or if the lesion is pre-malignant. No machine, no set of requirements, can circumvent these borderline cases and replace uncertainty with certainty. However, from a practical and clinical point of view, at some point in the evolutionary process, overt malignant behavior becomes manifest. This is accompanied by concrete and obvious differences between normal and malignant cells. These differences must be reflected in the patterns of normal cellular machinery expressed by the normal and cancer cells and their environments. [xx]

Cancer is characterized by abrupt discontinuities

Although the biochemical lesions of cancer can be subtle, and the molecular differences with normal cells can seem minor, cancer is not a smooth continuous process. Cancer is characterized by abrupt discontinuities. By analogy consider a pot of water heating on the stove. The temperature can smoothly increase. However, a small increase in temperature eventually leads to an abrupt, discontinuous change and causes the water to boil. To detect cancer cells we need to look not at the minor, subtle, continuous differences. Rather, we need to look at the major discontinuities. We need to see the boiling water. A cell is either dividing or not. A cell is either confined by the basement membrane or not. A cell is either invading surrounding tissue or not. A cell is either in the microenvironment of angiogenesis (new blood vessel formation) or not. The causes may be subtle, but these abrupt transitions are not. The discontinuities are dramatic and are reflected by equally dramatic changes in the local biochemistry. In other words, there are dramatic changes in the patterns of biomolecule expression or the patterns of protein expression that accompany malignant behavior.

The real test

The ultimate and only really meaningful test involves identifying a comprehensive set of target patterns and developing a real machine that can function in the same manner as our hypothetical cancer-curing machine. In other words, these requirements must be translated into the language of molecular biology, chemistry and drugs.

The theoretical requirements for the cure of cancer define a logical program and strategy

It follows on the basis of deductive logic from well established scientific theories and principles that the consistent and specific cure or control of cancer will require multiple drugs, administered in combination, targeted to abnormal patterns of normal cellular machinery that effect or reflect malignant behavior. A sufficient number of patterns must be targeted such that the probability of a malignant cell evolving without at least one pattern is clinically insignificant. (In practical terms, we estimate that approximately 5 to 10 drugs in combination will be required.)

The identification of an optimal set of target patterns and the development of a set of drugs to kill cells that express these target patterns is a solvable engineering problem. The technology exists today to develop a set of drugs that will function in exactly the same manner as our hypothetical cancer-curing machine.

Pattern Recognition Tumor Targeting (PRTT)

PRTT is a set of technologies to enable drugs that will kill cells if and only if the cells express a complete target pattern. Cells that express elements of the targeting pattern, but not the complete pattern will be spared. The elements of the targeting patterns can be inside the cell, on the cell surface, or in the microenvironment. The role of target patterns is to allow the detection of the malignant cells. The purpose of the patterns is to mark the malignant cells for destruction. The mechanisms of cancer cell killing are not related to inhibiting functions related to the patterns. The destruction of the cancer cells will be by independent mechanisms.

Any process that can consistently and specifically cure or control of cancer requires some type of PRTT.

Footnotes

[i] One bit is the information derived from a binary (yes or no) question in which the expected outcomes are equally probable to the receiver of the information.

[ii] The machine has no basis to expect that either a yes or no answer is more likely.

[iii] Few targets are actually present in such a high percentage of cancer cells.

[iv] An even better multiple-bit question has two parts, the first part asks about the potential for proliferation, the second part asks about invasiveness. This is discussed later.

[v] With respect to time, place and manner

[vi] The elements of the pattern can be on the cell, in the cell, or in the cell’s microenvironment.

[vii] Not just proteins, but any type of biomolecules can be involved.

[viii] In other words, a malignant cell could not realistically evolve without at least one of the patterns.

[ix] See the section on the logical implications of tumor cell evolution

[x] And the environment of malignant cells

[xi] Provided normal proliferative and invasive processes like wound healing are excluded as previously discussed

[xii] By definition, only cells that express malignant behavior are malignant.

[xiii] Some slight variations on these questions will also work. This technical point is discussed later.

[xiv] As a matter of logic, comprehensiveness cannot be proven. However, it can be refuted.

[xv] This is one reason why drugs targeted to EGFR have been so limited in their effectiveness against cancer.

[xvi] Nor is it possible to individualize the patterns to a given patient.

[xvii] Some cancer cells can evolve that actually thrive and grow better in the presence of anticancer drugs.

[xviii] The machine must also examine every cell’s microenvironment.

[ixx] The leukemias and lymphomas or liquid cancers will require a similar approach, but different target patterns.

[xx] The activation state of proteins is also important. There is a difference between MMP-2 and activated MMP-2. They are different chemical species.

[xxi] In other words,  "normal patterns of normal cellular machinery expressed in an abnormal context or setting"

References

[1] C. E. Shannon, "A mathematical theory of communication,'' Bell System Technical Journal, vol. 27, pp. 379-423 and 623-656, July and October, 1948
 
Andrews, F.C., Equilibrium Statistical Mechanics, 2nd Edition, 1975, John Wiley & Sons, Inc. P.64-71

[2] Duffy MJ.; “The urokinase plasminogen activator system: role in malignancy.”; Curr Pharm Des. 2004;10(1):39-49

[3] Wagner SN, Atkinson MJ, Wagner C, Hofler H, Schmitt M, Wilhelm O. “Sites of urokinase-type plasminogen activator expression and distribution of its receptor in the normal human kidney.”; Histochem Cell Biol. 1996 Jan;105(1):53-60

[4] Edelman GM, Gally JA.; “Degeneracy and complexity in biological systems.”; Proc Natl Acad Sci U S A. 2001 Nov 20;98(24):13763-8.