Predicting cancer prognosis ...
(It's been awhile since my last post, so I'd thought get a little techho here)
My Ph.D. project ...
In short, the goal is to figure out how to get a better handle on disease prognosis (i.e. predict outcome -- survival or death, remission or recurrence, how long someone might expect to be in remission, etc.). Specifically, I'm collaborating with a group that are experts in gastric (stomach) cancer, though we would hope our methods could be applied in other illnesses.
In the clinic, once it has been established that a patient has cancer, there are a number of indicators of how far the cancer has progressed. These indicators give doctors some idea of what treatment (or how harsh of a treatment) to use. But, these indicators are not very good for some cancers. It is unknown what patients respond to what treatments. And, gastric cancer is the second leading cause of cancer deaths worldwide.
From a panel of tumours collected over the last decade or so, for which clinical information is also collected, we are trying to find molecular "markers" (roughly speaking, gene or protein expression levels) of disease progression or of response to treatment. Of course, my angle is from the data analysis side, where we get the data, apply an appropriate statistical method and feed back candidate molecular markers back to the collaborators. They do further validation to determine whether what we find could make biological sense or eventually be used in a clinical setting.
There are a lot of challenges simply in processing and interpreting the data (some of which are fairly abstract), not to mention the challenges involved in validating a marker well enough to be used routinely in the course of treatment. The added potential benefit is that we can identify some interesting biology that also may lead to increased understanding and better treatments.
Should be challenging, fun, and hopefully useful some day.
posted by Mark Robinson at 9:16 AM
5 Comments:
notice that your stalking thread got more comments than this one?!
ANYWAY - this can be hupersonkind-shifting. best of canadian luck to you.
huperson = the new PC way of saying "human".
Saturday, July 16, 2005 6:21:00 AM
And, check this out. Cold War II?
China ready to use nuclear weapons against US over Taiwan: media
MARTIN PARRY of Agence France-Presse in Beijing
China could use nuclear weapons to retaliate against the United States if it attacked in any conflict over Taiwan, reports said on Friday citing a Chinese general.
"If the Americans draw their missiles and position-guided ammunition on to the target zone on China's territory, I think we will have to respond with nuclear weapons," said General Zhu Chenghu.
His comments were reported by the Financial Times and the Asian Wall Street Journal, which attended a briefing with the general organised by a private Hong Kong organisation, the Better Hong Kong Foundation.
"If the Americans are determined to interfere [then] we will be determined to respond," said Mr Zhu, a professor at China's National Defence University.
"We ... will prepare ourselves for the destruction of all of the cities east of Xian. Of course the Americans will have to be prepared that hundreds ... of cities will be destroyed by the Chinese."
Xian is an ancient city in central China.
Mr Zhu said the comments represented his personal view and not the policy of the government. Nonetheless, his threat to use nuclear weapons is the most specific by a senior Chinese official in nearly a decade.
The Chinese foreign ministry said it was "gathering information" on the issue and refused to comment further. The defence ministry also refused comment.
Analysts said the remarks were largely rhetoric but also indicated that Beijing wanted to show the United States it was serious about Taiwan.
"China's leaders, especially the military, think it is important for weaker countries like China to demonstrate their willingness to defend their core interests," Joseph Cheng, a political analyst at City University in Hong Kong, said.
"It's an old-fashioned strategy that is also meant for domestic consumption.
"On the Taiwan issue, the US government tends to adopt a strategic ambiguity approach. On the part of China, it wants to show it is ready to make the sacrifices," Mr Cheng said.
China's military spending has risen at an average double-digit rate over the past decade, hitting US$24.5 billion (HK$191 billion) in 2004.
Despite this, it would still not have the capability to fight a conventional war against the United States, leaving the nuclear card as the only option, analysts said.
Although China has a no first-strike nuclear policy, Mr Zhu said he believed the policy applied to non-nuclear powers and could be changed, the reports said.
China and Taiwan split in 1949 at the end of a civil war but Beijing still claims it as part its territory and has repeatedly threatened to invade if the island formalises its 56-year separation with a declaration of independence.
In March China adopted a law allowing it to use force against any secession moves by Taiwan, triggering concerns in Washington and raising tensions in the region.
Since the United States switched diplomatic recognition from Taipei to Beijing in 1979, Washington has acknowledged Beijing's position that Taiwan should be considered an integral part of China.
Yet the United States remains the leading arms supplier to Taiwan because it is bound by law to offer the island the means of self-defence if its security were threatened.
Mr Zhu's comments come ahead of Washington's annual report on the Chinese military and as a string of US officials have raised concerns about the rise of China's military.
The issue was most recently brought up by US Secretary of State Condoleezza Rice who, on a visit to Beijing last weekend, highlighted tensions with Taiwan as a key concern.
She urged Beijing to reach out to the island's elected leadership.
Saturday, July 16, 2005 6:31:00 AM
Ok, let's try and make a useful comment (I'll probably fail here).
1. Mark, never put in writing what you plan to do, and if forced be as vague as possible. It is good to have a plan, but you don't want to have it on paper. This was a) No one can hold you to it, b) If you end up making a great discovery accidentally, you can pretend like it was your great plan all along (Pastour is still regretting not listening to this last point)
2. Remembering of course that I know nothing about the topic, here is my question. Why not use the mouse model, where you can get a large number of mice, give them all stomach cancer (I'm assuming this can be done, if it cannot, then well who knows). Sepearate them into different degrees of progression, and then use a micro-array to look for different expressions in the different time frames?
This way, you get the benefit of seeing not only what expressions the tumors might produce, but also the expressions the body might make in reaction to them (which, from my point of view would seem just as useful, but again, I'm completely ignorant on the subject).
Of course, assuming you find some nice predictors, you then look for similar expressions in the human model.
(So, this might be a really bad set of observations and/or question, but I though at least one person should try and write something related to your thesis topic, so don't be too mean on your reply).
Saturday, July 16, 2005 10:55:00 AM
Steveo, yes, thanks for a relevant question ...
1. Yeah, I tried to be as vague as possible on the description ... and, as a matter of note, my boss really dislikes the type of research where people try all kinds of statistical methods, then choose the "best" one, then when they write it up, they say did ran the one they eventually used. A lot of people do it. Though, I guess that is a bit different than discovering things by accident ... it's more of a "code of practice" among statisticians issue.
2. Actually a good point about mouse models and some people are doing it this way. Definitely a more controlled environment ... problem comes when you try to clinically validate thing ... a scenario where "but it worked great in a controlled environment when all the mice were on the same diets, grew up with the same set of risk factors, etc". The guys I work with are actually right on the front lines treating patients as well as making biological discoveries. Which gives them unique access to human tumours of all races, genetic backgrounds, etc. If you can discover/validate markers there, then you have a better chance of promoting it to a clinical setting. Also, we do have normal stomach tissue to work with, both with adjacent normal tissue in the cancer patients and stomach tissue for non-diseased. (Even better would be a complementary approach of discovery in mouse, validation in human ... as you suggest.)
... a lot of hidden challenges that most current group do not even address, so there is lots of opportunity here.
Saturday, July 16, 2005 5:07:00 PM
just a note about mouse models:
1)we don't really have GREAT tumour models that resemble human cancers quite yet (although there are some really good ones)
2)unfortunately, we've been solving cancer in mice for a while depending on the model but the drugs aren't usually as good in the clinic. so if you can go human, go human. we just have to start tumour banking and gathering data better on patients
Tuesday, July 19, 2005 5:17:00 PM
Post a Comment
<< Home