Caution is warranted. Agreed.
Where are my #$%#$&!! manners?
Congratulations on the paper!
When uncertainty is prominent, many researchers regard Bayes’ Theorem as a way to make inferences and decisions. While I could have spent a few hours analyzing your Xenosite code published on bitbucket.org, it seemed more reasonable just to ask you: did you rely on Bayesian reasoning in this project? If so, how?
We use neural networks, which are designed to fit the posterior probability distribution (the prediction) of the mutation labels given their descriptors. This is a type of Bayesian method called MAP, which you can read more about here https://en.wikipedia.org/wiki/Maximum_a_posteriori_estimation and here https://www.cs.cmu.edu/afs/cs/academic/class/15782-f06/slides/bayesian.pdf)
We also use Expectation Maximization algorithm in conjunction with the neural network, to determine the labels. EM is a semi-bayesian method, https://en.wikipedia.org/wiki/Expectation–maximization_algorithm too.
Over all the method is definitely of the Bayesian school (rather than frequentist) but it is not a full Bayesian treatment.
Well that was a fun technical question =).
Let me add my congratulations here on your paper, which indeed looks likely to bear good fruit.
But I’m still having some trouble getting my head round the concept of “tumours evolving” rather than “populations of organisms evolving new susceptibilities to damage.” It sounds to me to be a metaphorical use of “evolution” that potentially functions to redefine mutations “deleterious to organisms” (the commonest type, because breaking stuff is easy) as “beneficial to tumours”.
It reminds me of Dawkins’ metaphor of the “selfish” gene as the true locus of evolution, which ticked all the population genetics boxes until the primacy of the cell in gene expression began to show it to be an inversion of the real situation.
Our evolutionary theories, after all, are about variation and selection in reproducing populations, leading to differential reproduction, are they not?. It’s not that tumours are not species - but that they are not true reproducing populations. Instead they are aberrations of cell-lines which even themselves are not loci of evolution, unless one has a very atomised view of eukaryote biology. After all, we don’t regard villous cells as “successful” because millions are produced daily, and Betz cells as “unsuccessful” because there are only 50,000 of them. We rightly see their role as subservient to the good of the true functional entity, the organism. How then can one rationally discuss the biological “fitness” of tumours, or even tumour-cells?
Since many BioLogos readers will have no medical background, let me briefly describe tumoursfor them. If a cell line - for example, melanocytes - has a particular genetic propensity, and is subject to the right environmental influences - for example, excessive UV - in just a small proportion of the susceptible individuals one, or a few, of the millions of melacocytes in the body will break free of the usual physiological regulation. If it also bypasses the body’s immune defence mechanisms, it may clone to form a malignant melanoma.
This is deleterious to the parent cell line of melanocytes, because they are destroyed by aberrant cells, deleterious to the organism (the true locus of variation), and ultimately deleterious to the wildly reproducing tumour cells, because they die with the patient… though presumably neutral theory correctly predicts how and why the particular genetic markers involved exist in the human population.
If one may treat tumours as a quasi-species “benefiting” from mutations, and so evolving, then equally one can talk of immune cells “evolving” because genetic patterns of auto-immune disease cause them to proliferate. But is that not ultimately a misleading view down the wrong end of the telescope?
So that’s my first problem. My second is that in rejecting ID arguments regarding evidence for design (not sure which ones - is it the alleged rarity of beneficial mutations, in which case the validity of “tumours” as a true locus of evolution becomes crucial?) you have also apparently rejected any actual divine involvement:
…or regularly require the direct intervention of God to initiate and be sustained. I’m not sure which option is harder to believe.
That seems to be a theological, not a scientific, statement. It always plays well to deny the involvement of God in anything harmful (though of course in the Bible God does actively send tumours, albeit bubonic rather than malignant ones), but the question needs to be argued, not assumed. If it is hard to believe in principle that God (or why not the devil, if one is a believer in fallen creation?) would be associated with these genetic changes, then one is saying that they are excluded from God’s providence.
More specifically, in the context of the belief called “Evolutionary Creation”, in which “evolution is God’s means of creating life”, one is saying that not all evolution is God’s creation, but only some.
In which case, one needs to state clearly what other “demiurge” has its muddy paws in the process of evolutionary creation, and presumably be able to have some emprical way of discerning which parts of creation are designed by God, and which by this other agent (chance, maybe?). That would, it seems to me, only be a particularly subjectively discriminatory form of Intelligent Design.
Good questions Jon! Thanks for the note…
I see your objection and am sympathetic because I held it for a long time too. It is easy to dismiss all we have learned about cancer and its relation to evolution as an example of things “breaking down,” but look at this from the point of view of cancer cells “evolving” is perhaps the dominant paradigm, and is a more accurate approach for a few reasons.
Cancer cells are not “less” functional, as if they were broken. In fact malignancy requires several (some estimate up to 15) often convergent genetic changes (i.e. angiogenesis, cell-adhesion, etc.). The preferred term is “disfunction” because it recognizes correctly that cancers are functioning inproperly (not failing to function), and this is often achieved with neofunctionalization,at a molecular level, something that many ID people think is impossible. This fits the evolution paradigm better the broken paradigm.
Cancer cells clealry adapt to their environment and this is fundamentally important to treatment, because they develop resistance to drugs by evolutionary processes.
As a model system for evolution, it is particularly useful because the genetics of cancer match humans better than any other model. There are whole classes of mutational systems (e.g. splicing variants) that can cause cancer but are not really at play in microbial systems.
All the mathematical machinery of population genetics and neutral theory seem to work exactly as we expect they would on cancer. So perhaps it is an analogy (because cancer is not an independent organism). Though, some versions of cancer do function like transmissible parasites (https://en.wikipedia.org/wiki/Canine_transmissible_venereal_tumor) but that is atypical.
Finally, the reasoning we use is identical to evolutionary reasoning. We use data in the present to make inferrential statements about the past.
Now, you are are right, cancer is not exactly evolution, because new species are not evolving. That is where the analogy breaks down. I, rather, think of each tumor as its own independent evolution experiment (and also a tragedy for the patient that we want to cure).
I think you are missing what I meaning. What I am saying is different.
If many ID arguments about neofunctionalization and molecular convergence being impossible are true (as many ID proponents are convinced), how do we explain neofunctionalization and molecular convergence in cancer? Clearly it is the same biological system, and we are seeing neofunctionalization and molecular convergence, and all the same genetic patterns in cancer tumors that we see in, for example, humans the the great apes.
In the case of evolution, we look at this data and some ID proponents conclude that it must have required God’s (thought they would say “designer”) direct intervention somehow (from tinkering to special creation).
In the case of cancer, the data looks very very similar. If the ID proponents are right aboutevolution, why would we conclude anything different here?
To take the theology out of it, i should have said,
or regularly require the direct intervention of [an intelligent designer] to initiate and be sustained. I’m not sure which option is harder to believe.
The point is not about theology (though it certainly raises theological questions) but is a question of how the ID scientific logic is applied in two places (evolution and cancer) where we have nearly identical genetic data.
Good explanation. In layman’s terms, this is why another round of the same chemo that did well the first time, will not work as well the second, and why recurrent cancer is tougher than the primary. You can kill the easy cells, but the more resistant ones take over and grow.
Variation, propagation, selection and inheritance: These are the bases of biological evolution.
Cancer cells are clones. Much like bacterial cells form clonal colonies on agar, tumors are colonies formed from single cancer cells. As these cells replicate, additional variations accumulate, leading to further diversification of the population in the tumors. Over time, cells may separate and found new tumors elsewhere. Some of the changes acquired by subsequent generations of cancer cells support better growth, others, better propagation or bypassing of self-destruct signals. Vascularization, penetration though tissue boundaries and resistance to chemotherapy agents can also be acquired and honed by selection. The cells in say, later stage cancers, often have accumulated new traits compared to the cells found in earlier stages.
It is perfectly reasonable to approach cancer as a providing a lesson in evolution. That is why, for example, when we evaluate new therapeutics, we also try to identify both the cancer cell populations that are likely to respond and the probable pathways to resistance. In many cases, the targeted therapies work only for a limited duration. We can often knock down the more populous, susceptible cells and achieve a term of remission but resistant clones frequently take over and drive the next emergence of the disease.
Also, I found a few open access journal issues for those of you wanting to see more clealry the relationship between cancer and evolution. These are good reads.
So very interestingly this thread was picked up by the ID movement at UD and Dr. Rossiter’s blog…
I hope that some of us join us for the conversation here. If they do, please do treat them with respect.
And now ENV has decided to jump in.
I had no idea that they followed this thread! Apparently those here are noted as a “jocular” biologos crew. Interesting.
And this one too…
Kinda missing the point entirely…
And yet another one…
I guess I am flattered that people care what I think so much. I should also positively note that (for the most part) the ID people are not resorting to many ad hominems against me. I do appreciate that. Though I do agree with @Argon
[quote=“Jon_Garvey, post:18, topic:5673”]
But I’m still having some trouble getting my head round the concept of “tumours evolving” rather than “populations of organisms evolving new susceptibilities to damage.”[/quote]
Are you saying that the term “evolution” can only be applied to whole organisms?
If proteins evolve, why can’t tumors?
[quote] It sounds to me to be a metaphorical use of “evolution” that potentially functions to redefine mutations “deleterious to organisms” (the commonest type, because breaking stuff is easy) as “beneficial to tumours”
[/quote]Except that the other common type of tumorigenic mutation is gain-of-function, not “breaking stuff.” There are many fascinating fusion proteins as well.
Have you heard of oncogenes?
Wow. Ann Gauger recites the two classes (oncogenes and tumor suppressor genes) ***, but then discusses only p53 (a tumor suppressor gene) as an example of a gain-of-function oncogene!
*** (content deleted by moderator for lack of graciousness)
It’s fascinating that they dropped your MD for the title. I suspect that Gauger wrote this one too.
ENV:As to the prediction that oncogenes accumulate point mutations to become cancer drivers, that’s interesting, but from reading the paper one sees no discussion of any specific point mutations…
This is just laughably false. The paper is full of discussions of large SETS of specific point mutations. Moreover, when real scientists read a paper, we know that most of the information is in the figures, tables, and online supplements, not in the text. It’s clear that the ENV author didn’t bother to really read the paper before writing about it.
or what their specific phenotypic effect is in any specific genes.
That’s not what the paper’s about. It’s written primarily for people who are already familiar with the genes discussed, unlike the ENV writer. And what does phenotypic effect IN specific genes even mean?
If Swamidass knows of examples that show “cancer regularly innovates with proteins of novel function,” his paper discusses none of those examples.
Of course not! That’s not the point of the paper. It’s already very well known. The paper was about identifying those and other important genetic changes.
They also leave out that I am a computational biologist with expertise in information theory, and a professor. The article text does clarify some. I think they maybe are just choosing the least relevant title to this conversation in the title.
That is okay though. I certainly am a computer scientist and this isn’t about titles.
I am glad that you called him out on this cheap shot:
“But he has been relentless in his assault on all views that allow for a Christian God to be directly active in the creation and evolution of biological life. In specific, he has targeted ID forms of those arguments, while entirely belittling those views considered more biblically conservative.”
To demonstrate that I have been very clearly not been doing this, can you elaborate my position a bit for the newcomers @Jay313? Perhaps include some links to me defending YEC’s in other threads =).
I think in the meantime it is very important to correct this misstatement of my position: “But he has been relentless in his assault on all views that allow for a Christian God to be directly active in the creation and evolution of biological life. ”
I have NEVER made that case. I absolutely believe that the Christian God has been thoroughly active in the creation and evolution of biological life. I have never argued against this belief. I am not a deist in disguise. I am Christian that believes in God’s sovereign will over all things, including the creation of us all.
Moreover, I have bent-over backward to acknowledge the possibility of the special creation of man. http://peacefulscience.org/category/projects/100-year-old-tree/
You write, “while entirely belittling those views considered more biblically conservative.” I would love to know (please send me a private email) where I have done this. I consider my point of view very biblically conservative, very much in the line of Martin Luther and Blaise Pascal.
I should emphasize that I certainly think that YEC can be a respectable position, and regularly refer to Kurt Wise and Todd Woods (and more recently John Sanford) as YEC’s that I respect. I also have a great deal of respect for the folks at RTB (including Hugh Ross, Jeff Zeerwink, Fuz Rana). Of course, I totally disagree with aspects of their science. But how does voicing and explaining these scientific disagreements constitute belittling their biblical views?