Watson in Jeopardy

From time to time a new computer, robot or piece of software comes by and impresses us so much that I think it cannot be real, must be a fake of some kind.
When Deep Blue faced Kasparov in a series of chess matches, winning 2-1, it is reported that after a particularly unexpected move, the human player stood up and accused the computer of being controlled by humans: a computer could never conceive such a creative move.
When I watched the recent Jeopardy match between the computer (robot? software?) Watson and two champions of the game, I had the same feeling. The first thing that impressed me was that the moderator of the game spoke so fast that the algorithm used to translate the speech into text had to be really great.
Second, translating speech to text is one thing, now how to place this into context? At certain moments the moderator would address Watson and ask questions while in other moments the moderator would talk to the other participants or even the audience, so how did Watson know he was being addressed? There were probably keywords that are used in this game, and the programmers that built Watson probably added these as reference, but by what I saw, in only one occasion Watson spoke out of time, when he incorrectly pronounced the detergent "Cheer" as "Cheering".
Third, the questions of this game are a challenge by itself. I noticed that the two human players took at least one attempt before they understood what the category of questions was about, while Watson didn't miss any of them. How could he "understand" something better than a human? I will reinforce the question: humans have this belief that we are better capable than anything else on Earth of understanding a situation, making sense of it. After all, we can watch tv, which is a projection of a 3D universe into a plane. We can watch a six year old drawing of car and understanding that it means a car. So how come Watson was capable catching the ideas of the questions so fast that the other players would look at each other with that "Ahan..." look?
Fourth, once the scope of the question was defined, how did Watson translate the meaning of the question to the answer? We can conceive that Watson has almost limitless storage capacity and a super-fast processing, but how do you determine that George Bush rhymes with "tush"? Does a computer know what rhyme is? Does Watson have stored somewhere the notion of rhyme, or maybe of sound?
I don't believe so, it appears to me too much, but perhaps I am wrong. What I believe, however is that perhaps if you are fast enough and have enough processing power, you can come up with solutions that were not built with logic reasoning, step by step, as we do. Maybe Watson is not a robot built to think like us, maybe he was not even built to think in first place, perhaps he's just like the room full of monkeys which eventually write down all the work from Shakespeare.
But this rises at least one question? Even if Watson is quick and powerful enough to create a huge number of different answers almost instantly, he still needs to parse those and choose only one, and this implies that he must be capable of comparing the answers and saying which ones are better than the others. Doesn't this imply that he has at least to understand the question?
This hypothetical approach would be similar to a person who cannot solve a differential equation but who can derive expressions, thus he can generate a random list of functions and test each of them as the possible solution for the problem. But again, how does Watson score his answers?
Of course other possibilities for what we saw in Jeopardy, for instance, it is possible that IBM had a thousand different Jeopardy experts playing in a network and they all submitted their answers within the 3 second limit, the answer most voted would then be chosen, but I don't believe this happened. The answer that Watson gave to a tennis term which was also used in races was so out of contest ("automobile races" instead of "rally") that it rose giggles from the participants and audience.
A second possibility is that Watson is connected to a Doomsday machine: every time Watson gave a wrong answer, the universe was destroyed (or at least Earth, or at least me and you reader). In universes where Watson got the right answer we would be left baffled by this wondrous machine. I first read this example of the ultimate computation in a book by Hans Moravek "Mind Children", and of course would only work if the theory of parallel universes is valid.

Can the knowledge of the observer alter the subject of observation?

There are at least two well known novels where a character is capable of "seeing" the future, the first one is Foundation from Isaac Asimov and the second is Dune from Frank Herbert.
In Foundation, a mathematician name Hari Seldon develops a branch of Mathematics named PsychoHistory which he uses to model the universe (more specifically the interactions across a galaxy-wide humanity) and predict how it would behave. Hari discovers that the galactic empire is about to fall (within the next hundreds of years) and he uses PsychoHistory to interfere with this fall in order to minimize it and create a Foundation for the creation of a new empire.
In Dune, Paul Muad'Dib, the Kwisatz Haderach, has the power to see the future through trances induced by Spice, a mysterious drug. In Dune, many are able to see hints of the future thanks to the spice but Paul is the only one that can clearly see the future of human kind, much like Hari Seldon, as much as his own.
Both are tragic characters, like Cassandra from Greek mythology, and their gift of vision into the future eventually becomes their ruin.
These two examples along with memories from my Logic classes and Goedel Incompleteness Theorem (basically a system of rules cannot contain itself and thus cannot self-validate) made me think that any observer could be classified into at least three categories: (a) An observer with little knowledge (predictive power) of the subject being observed, (b) an observer with total knowledge of the object and (c) an observer with knowledge of the object and itself.
In the first case the object's properties are independent of the observer's existence and thus the observer may find the subject "unpredictable" or predictable but showing some "noise" or "error".
In the second case the observer knows in advance the next state of the subject of observation. Considering that this is known ahead of time only to the observer, does it imply that the object is determined by the observer? Does the observer see the future or does he create it?
The third case is the most elaborate: the observer not only knows how the subject will behave but also how he (the observer) will react to this and thus the subject is aware of an immutable future in which he is trapped.
Seldon and Muad'Dib also had another ace in their sleeves: both were leaders of men and eventually were believed as Oracles, Seldon after death, MuadDib in life. Seldon continued to guide his followers through video recordings that explained the current state of matters and how they should act. Was Seldon using these communications as a way of shaping the future? Were his predictions nothing more than tricks that made people believe that he could actually see the future and thus they would do as he said, thus creating a future foreseen by him?
Muad'Dib becomes emperor of the known universe with the goal of avoiding a great tragedy that would eradicate mankind. Eventually he becomes the cruelest tyrant in History in his desperate attempt to steer mankind into the right direction. No matter how much he tried, the universe found a way of slipping through his fingers and returning to his apocalyptic visions. Eventually he realizes that his own existence and his influence in others was the key log that kept that terrible future immutable and he decides to walk into the desert alone towards death.
So comes the final question: do we have really free will or are we only deluded because of our lack of knowledge of the universe and ourselves?

Slums in big cities and tumor invasion

In this assay we compare two phenomena: tumorigenesis and the development of slums in big cities, and propose that not only the rules that control their existence are similar but also that the strategies in order to eradicate them are equivalent and that the lessons learned from one problem can be used in the other.

Slums are a grave problem in big cities in underdeveloped and in development countries. In 2007 in São Paulo, the biggest city in Brazil, there were approximately 2,000 slums with a total population of more than 400,000 families living in sub-human conditions. Besides the social problem of this population deprived of minimum sanitary conditions, slums are also safe haven for organized crime and drug dealers and gradually grow by engulfing neighborhoods of the city whose real-state is downgraded by the proximity with them.

Tumors are believed to be created by the relentless replication of genetically unstable cells that, through mutations and selection from microenvironment, acquire a set of phenotypes that allow them to invade healthy tissue, promote angiogenesis and colonize new regions of the host and create new tumors [1, 2], eventually reaching a state of tumor burden that is fatal to the host.

Both phenomena, slums and tumors, often develop in the periphery of the host (carcinomas develop from epithelial tissue separated by host by basement membrane while slums have their origin in the outskirts of towns where real state is less expensive) where resources are limited and uncontrolled growth lead to gradients of resources and harsh conditions.

Both systems invade by “trashing” their surroundings: tumors invade healthy tissue by both degradation of extracellular matrix and by causing death of healthy cells; it is known that tumors constitutively metabolize glucose anaerobically producing lactic acid [3, 4] even in presence of oxygen. It was proposed that this glycolytic phenotype would be a mechanism through which tumors would intoxicate their surroundings in order to kill healthy tissue and make room for new tumor cells [4]. A similar mechanism is found in the periphery of growing slums: a wave of devaluation of real state moves outwards of the slum propagated by criminality which imposes a “bad reputation” to the neighborhood, scaring the dwellers away and leaving room for new residents from the slum periphery or from outside of the system.

Solid tumors are often avascular during the early steps of tumorigenesis and are only able to promote angiogenesis as they achieve a critical mass. The fragile infrastructure of slums is no different from solid tumors: as one progresses into the settlement, the roads become narrower until cars cannot traffic, what considerably reduces efficiency of law enforcement. This inability of law enforcement and a minimum infrastructure for the survival of the slums is similar to what happens in solid tumors. In one side poor perfusion prevents a faster growth of tumor but on the other hand it protects the tumor by preventing the action of the immune system, chemotherapy and radiotherapy by limiting diffusion of drug, inducing quiescence in hypoxic tumor cells and by generating a heterogeneous microenvironment that confers robustness to attack [5].

We have discussed some aspects in how carcinomas and slums develop in a similar manner, notably by uncontrolled population growth in an area in the edge of the host/city with poor infrastructure but also with small or no interference from immune system/law enforcement, as is the case with carcinomas which are separate from immune system by a basement membrane.

Both systems appear to be robust to brute force attacks (toxins and antibiotics in cancer, and law enforcement and eviction in slums) not only because these approaches cause higher side effects in the “host” than in the target but also because the forces that promoted the initial development of these systems remain unchanged (genetic instability and microenvironment-imposed selection for cancer, and social inequality in slums) and thus will promote regrowth of the original system or other similar ones in other areas.

We propose that the most promising strategies for eradicating and preventing carcinomas and slums are those that target the forces that promote their emergence. For carcinomas these strategies would focus on intratumoral pH normalization, use of glucose competitors, use minimum amounts of therapy necessary to arrest tumor growth and delay patient relapse, and finally assess tumor response to therapy in a closed-loop approach. For slums, whose emergence is due to a considerable mass of poor people, the most promising approach would be to invest resources into bringing this share of the society into more equal conditions, which can be achieved by full-time public education with meals and recreational activities in order to keep the children away from one environment permeated by violence, drugs and poverty. Work laws that ensure minimum wages and social programs to provide credit to families to finance homes are also more immediate actions. Finally, the problem of slums in big cities will never be solved if the flow of migrants from poorer underdeveloped regions of the country remains. It is important thus that such an action for reduction of social disparities happens country-wide.

As a final note, we would like to stress that even though slums carry within criminality and major social and public health problems, they only exist and grow because of the initial advantage of cheap labor they offer to the richer population of the cities. An interesting point is that in carcinomas the cells that develop as tumors are exactly those that are isolated in the periphery of the host and considered as “expendable”.

1. Goldie JH: Drug resistance in cancer: a perspective. Cancer Metastasis Rev 2001, 20:63-68.
2. Hanahan D, Weinberg RA: The hallmarks of cancer. Cell 2000, 100:57-70.
3. Gatenby RA, Gillies RJ: A microenvironmental model of carcinogenesis. Nat Rev Cancer 2008, 8:56-61.
4. Gatenby RA, Gawlinski ET, Gmitro AF, Kaylor B, Gillies RJ: Acid-mediated tumor invasion: a multidisciplinary study. Cancer Res 2006, 66:5216-5223.
5. Kitano H: Cancer robustness: tumour tactics. Nature 2003, 426:125.

Mets are Vets?

The other day I was talking to one of my advisers and he mentioned that a reviewer had just critiques the fact that the Acid Mediated Tumor Invasion does not consider the fact that the aggressive cells selected by tumor microenvironment (hypoxic and acidic) should lose this phenotype once they progress through healthy tumors, since the selective pressure on these cells decreases as they invade further from original tumor.
A metaphor used by that time was that tumor cells are like Navy Seals who, once far from their training environment, relaxing in a paradisaical beach would eventually get soft like other regular people.
The point here, I believe, is that the metaphor is wrong for tumor cells are not trained to be tough, they are selected, they are traumatized like 17-year old boys who are sent to war. Those who are able to adapt, to toughen up, survive and come back home, but many of them cannot adapt back to normal life.
The perfect example is John Rambo. When Rambo, and many other Vietnam war vets, returned to US, he no longer could live in society as we do. He would see threats everywhere, he would trust no one and would not accept the authority of civilians. In the movie Rambo escapes from prison and hides in the woods nearby the city. When the police and National Guard try and capture him he turns back to the Green Beret M.O. and kills them all.
At the end of the movie he destroy much of the city including the Police Station.
I believe that the cells that survived the stress created by tumor development and reached a more healthy environment to live will, in their majority, settle down. The cells that start new metastases however are like John Rambo: they cannot re-adapt to normal environment so they tend to re-create the original environment that traumatized them.
If this is true, there may be ways of acquiring an aggressive phenotype through reversible ways (most of tumor cells) but some of these ways may yield to irreversible conversion and eventually lead to cells that can generate metastases.
A microarray analysis as well as protein expression of cells recovered from metastases compared to the original tumor cells might yield to clues on how this commitment happens.

Substrate-induced starvation in cells?

I was just reading an article sent to me by my PhD adviser about evolution of metabolic networks.
It is impressive how metabolism is structured in a complex network with apparent redundancies, inefficiencies and bottlenecks.
But this post is about something extremely interesting I read in one of the references: apparently some of these metabolic pathways are "boosted up" by an energy-dependent step. These are called "Turbo-designed pathways" and glycolysis is one of them.
In glycolysis, before glucose catabolism can yield 4 molecules of ATP, first it consumes 2 molecules.
When yeast is grown in culture with high glucose concentration, negative feedback mechanisms prevent the cell from wasting too much energy in accumulating substrates intracellularly and at the cost of a lot of energy. Yeast just uptakes enough glucose, phosphorilates it (2 moles of ATP consumed) and then metabolizes it to get enough energy.
However, when yeast is grown in media with high malate (analogous to glucose) concentration, these cells will deplete all their energy uptaking this substrate and before they are able to make out the ATP surplus, they are already dead.
It is like someone spending all their money in seeds and not saving anything for sowing the crops.

The funny thing is that tumor cells are known for increased glucose metabolism, not only increased glucose consumption but anaerobic glucose metabolism, which is much less efficient. So wouldn't these cells just starve to death? I guess they would, if the increase in the metabolic flux where in the ATP consuming steps, what would lead to accumulation of metabolites and energy deprivation. But if the increase is in the ATP generating steps (after glyceraldehyde-3-phosphate dehydrogenase) then it would yield to energy surplus.

This conclusion is very interesting because it suggests that any increase in expression of glycolytic enzymes pre-GAPD should come only after the ATP-surplus enzymes have already been mutated (or its expression increased) to increase its reactions fluxes.

For references:

The danger of metabolic pathways with turbo design. Teusink et al., 1998
Glycolysis, turbo design and the endocrine pancreatic β cell. Iynedjian, 1998

Humankind and cancer

One Day I was talking to my postdoc mentor about how a therapy for total cancer eradication should look like. We were not arguing any specific strategy bust just were trying to imagine how a cure for cancer would look like and how we would describe it if we found a Djinn in a desert.
The dialog proceeded a bit like this:

Mentor-An average tumor should have something around 5-10 billion cells. This tumor, however, is often spreaded around the body in regions from which it cannot be extracted without causing patient death.
I-But can't we help the body fighting it? Cancer immunotherapy proposes to identify over-expressed proteins that can be targeted by immune system.
M-Do you know the success rate of these therapies?
I-5-10%?
M-You'd be surprised by what a placebo can do...
I-So how could we envisage a therapy? Any sort of cure? Maybe keeping cancer as a chronic disease, just slow down its growth?
M-Coming back to the tumor size, imagine humankind where a tumor on earth, there are around 6 billion people, spreaded all over the world. They eat differently, have different resistant to diseases, live in different weathers, humankind is pretty much as heterogeneous as tumors...and as pernicious to the planet as tumors are to human body.
I-So the problems are equivalent: how to eradicate human kind and save the planet is like eradicating a tumor?

Now how can we overcome such a challenge? The more we proliferate on Earth and the more we pollute, destroy natural ecosystems, the closer we get to our own extinction, but will life on the planet survive our extinction or shall we bring all life with us like a real cancer?

Why do tumor cells glycolyse?

Since Pasteur and Warburg it is known that tissue tends to increase glucose uptake in low oxygen environments and that tumors continue metabolizing glucose anaerobically even in presence of oxygen.
Warburg originally hypothesized that this effect not only was a hallmark of cancer but also that is was its main cause. Warburg believed that all tumor cells had malfunctioning mitochondria and thus were forced to rely on anaerobic energy metabolism.
Gatenby and Gillies recently proposed the acid mediated tumor invasion hypothesis, through which, not only glycolysis was necessary for tumor proliferation and invasion, but also that this phenotype is selected by the micronenvironment were epithelial tumors are formed.
This hypothesis proposes that the use of abundant glucose in serum anaerobically would provide enough energy for tumors to develop and also produce acicity that promotes degradation of extracellular matrix and induces apoptose in normal surrounding tissue.
One interesting point, however is that tumor do not entirely abandon aerobic metabolism. Oxygen is necessary for keeping the basic functioning of TCA cycle that not only produces energy but also is linked to production of fatty acids and and many other metabolic pathways in cells. It has been shown that tumor regions in anoxic conditions become too acid for even tumor cells to survive.
I propose that the reduction of O2 consumption in tumors is an adaptation at population level. The cells that are in tumor rim, closer to oxygen-rich tissue, consume less oxygen which diffuses through them towards the cells in the inner core of tumor, preventing excessive acidosis and allowing tumor growth.
This kind of behavior is not normally proposed in tumors, that are normally considered as a heterogenous disorganized mass, but maybe these systems of cooperation between different subpopulations in a heterogeneous environment are exactly what differentiate malignant tumors from the many neoplasias that are periodically destroyed in our bodies without our knowing.

Democracy and Cancer

For long time it has been known that molecular signaling in cells is done through scale-free networks, both at gene and protein levels.
Such networks possess an architecture that confers resistance to failures but fragility to attacks.
The Internet uses such architecture: every computer is connected to an "internet access provider" which is connected to other servers that are interconnected all over the world. The design of internet addresses is an example of this hierarchical architecture, where each node is directly connected to approximately the same number of nodes (or at least the same order of magnitude).
A failure in a random computer in the network is not likely to affect the network performance since the chances of this computer being in a node in the lowest levels of the network (a user computer) is a lot lower than being one of the servers. Also, the catastrophic consequences of a failure in one main node in the network, associated with the low number of computers in this level, allow spending more money in protecting these high-level nodes with backup systems.
This architecture also allows control of network flux (it is sufficient to control a few high level nodes in order to control most of network flow) but also raise the risk of attacks.
Many examples exist of hacker attacks directed to websites, corporative networks and even entire countries.
How does this links to cancer?
It is widely accepted that cancer is originated from genetic instability. Many other factors, such as population heterogeneity (also born from genetic instability) and environmental conditions (selective forces, inflammation, exposure of body to carcinogens, etc.) are also important and may harness the genetic variability and direct towards malignancy.
Genetic instability can be considered as failures in the regulatory networks of the cell. These failures are most probable to happen in nodes in the lower levels of the network, for these are the most numerous. The mutations can occur in three ways: the first would be either a silent mutation or a mutation that would cause little effect on cell state, the second being a mutation that would alter cell state in a way that it would activate cell programmed death mechanism.
The third and most dangerous way if a mutation that alters cell state but bypasses apoptosis. Such mutations, if accumulated, would allow cell to escape body's control system and start a tumor.

Another point I'd like to point out is that the scale-free architecture from cell networks are optimized for evolutionary purposes: when a node is connected to many different others, and each of these others is connected to many others, it is possible for this system to gradually evolve to selection through increase of weight of influence from one of the 2nd degree nodes but also show more dramatic evolution due to modification of weight of a 1st degree neighbor or of the node itself.

Lastly, this architecture is very similar to a democracy, where most of the people has one low value vote, while fewer have higher level votes (congressmen) and very few have rights to alter the state of entire groups (governors, president). The idea of democracy is not that the majority is smarter than individuals but rather that a great number of people are less likely to take decisions that will harm the entire group. Due to practicity, democracy had to be stratified to allow decisions to be taken more rapidly, and thus appeared the scale free network of decision making known today is most democratic nations.

Most of the efforts today in cancer treatment are focused on finding targets that can be treated by highly specific drugs. It is believed that if these targets, which are often mutated in cancer, are silenced, cancer cells would die or at least lose its advantage over normal cells.
This idea would be correct if the only mutated nodes were the ones selected by the therapy. Unfortunately cancer is not based on attack but rather to failure and thus the probability is that, if a high level node has been mutated, this event was preceded by many more other mutations in intermediate and lower levels nodes all over cell signaling network.

Instead of a corrupt governor or president, cancer is more likely to be the result of a coup d'êtat that gradually formed from the infiltration of enemies all over the hierarchy of cell network. If one leader is removed from the network, it is likely that many other still exist in a lower level and will eventually replace the lost comrade.

In such a situation, the strategies to be used would be similar to the ones that have had success in fighting corruption in poor countries: the development of education and the increase of level of life of citizens tends to reduce the chances that corrupt leaders will be elected. The same is true for cells: if the micro environment of tumors is changed to endanger the survival of cells, there will be an opportunity for these cells to suffer mutations and diverge towards cancer. Conversely, modification in micro environment may reverse this tendency by preventing new cells from turning into cancer cells and also by reducing the efficiency of the tumor phenotype which is adapted to adverse conditions.


Regulatory network of genes with known role in cancer.
From "A map of human cancer signaling."
Qinghua Cui, Yun Ma, Maria Jaramillo, Hamza Bari, Arif Awan, Song Yang, Simo Zhang, Lixue Liu, Meng Lu, Maureen O'Connor-McCourt, Enrico O Purisima & Edwin Wang doi:10.1038/msb4100200