Hi Everyone,
I decided to present the following scientific article on tackling the problem of drug-resistant bacterias (commonly known as superbugs):
http://nautil.us/issue/43/heroes/will-viruses-save-us-from-superbugs
This is a nice read.
To cut a long story short it talks about (hopeful) renaissance of well known viruses that kill bacterias, especially the antibiotics resistant ones. In order to prevent from being killed the bacteria has to evolve, and by mutating it is losing a drug resistance becoming vulnerable to antibiotics.
There are questions that pop up in my mind and I would love to see your answers to them:
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
Thanks and Best Regards,
Paweł.
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeletePerhaps we could perform multiple simulations of bacterias evolving and simulate any new viruses behavior. If the number of such simulations was sufficiently large it would allow us perhaps to discover some useful findings.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
As mentioned above, I would start off with defining all possible models with random variables and try to simulate its bahavior.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
The task could be perhaps handled by unsupervised or reinforced learning.
Thank you for your insightful comment. I thought about something else, but simulations actually make sense, thanks.
DeleteHello, thanks for sharing this interesting article with the rest of us.
ReplyDeleteLet me get straight to your questions:
1) I guess the best way to increase the speed of finding matching viruses is either by conducting more experiments or trying to categorise and find viruses of the same genus, family? I don't really know how they are categorized but I guess they could by showing similar traits and characteristics. One thing I would be careful with is to check if said virus isn't harmful to other living organisms that are potentially beneficial for us.
2) As Ivona already mentioned a simulation would work wonders in this case.
3) Unfortunately I am uncertain of which would fit here the best.
Thanks, Michail.
DeleteI must agree with aggregating viruses by types of genomes, there is got to be something in common that make them suitable to be used against certain bacteria species.
Perhaps increase the speed of tests would help as well. It might be done by some degree of automation. However, to make results useful, I believe we need some kind of database...
Paweł, thank you very much for this great and exciting article (and none tech ;-). I always try to read article about the latest medicine revelations because in my opinion health this is most important issue in human life. This threat method which has been described in this article is very interesting, hopeful but in my opinion always dangerous. I reckon that all tricks with nature (viruses and batteries) are very dangerous because their influence and development are unpredictable. Today some issues can help us but tomorrow the same issues can kill us. Moreover I can only hope that all governments will open their eyes and minds to a possible cure for so many incurable illnesses and fund the research on this. We seem to have some closed minds if money is not to be made. Unfortunately money above all else seems to be pharmaceutical company motto.
ReplyDeleteThank you for your reply.
DeleteThe article is still on science. What I wanted here, was for us to brainstorm on technical solutions that would help make this world better (in this case by eliminating drug-resistant bacteria species).
I also thought that using viruses to kill bacteria could be quite dangerous and we might face unforeseen consequences. However, if you think about it, it seems that no new virus strains are engineered here. We want to use something that we already have in the ecosystem and it does not affect humans. Or at least we are not aware of negative interactions...
Nevertheless, funding research on such a high-stake field is always a good idea.
This is a great topic indeed. I didn't know something like this even exist. But I'm afraid this is completely new subject for me and I'm feeling totally unable to put something into the discussion. But I'll try.
ReplyDelete1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
Only general form of an answer comes to my mind - and yes, it was already mentioned few timnes - performing simulations. But what exactly should be simlated, where simulations should be taken and by whom simulations should be conducted - don't ask me.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
I really have no idea. I have an experience that if you'd like to use some software, or create one, you have to do the research first. I didn't do this kind of research yet for that topic. Maybe it should be some simulation software which will let the user check which phages can turn which bacteries into "phage-copying machines" as mentioned in the article
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
I don't know which would be the most promising, really.
Thanks, Paweł.
DeleteHonestly, I had something in mind. I am a bit fixated on Deep Learning, so Ivona's comment opened my eyes a bit.
The way I understand how it works (given my limited knowledge on biology it might be way off) it should be possible.
The author mentioned that certain virus strains match specific bacteria species. From what I read, I think that it a virus enters the bacteria through the mechanism it uses to protect against antibiotics. Human cells use the same tactic to get rid of toxins.
Going back to the mechanism, the bacteria specimen will mutate as a response to virus infection. But that makes it vulnerable to antibiotics. It seems like a win-win situation. If we can create an algorithm that will find the right virus to match the bacteria protective mechanism, we are all set.
There are quite a lot of algorithms that can search the large problem space (i.e. evolutionary algorithms like genetic algorithms). Currently there is a lot of buzz on Deep Learning, which in my opinion is suitable here.
The topic is very interesting, but I do not know about it a lot, so comment fully I can not.
ReplyDeleteThanks Maciej, nonetheless. You sound a bit like a master Yoda (not that I know anything about Star Wars, but...)
Delete1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteWell the only thing that comes to my mind is to use computers. We may create some database with all antibiotics and other medical treatments with diseases and bacterias that they cure as well as database with bacterias and all informations that they collected about them. We may use machines to try all possible connections of antibiotics and bacterias to see what might work.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
There is bioinformatics and other fields of education that focuses on medical usage of computer science. I am not an expert in this field, but for sure there are softwares used right now, but maybe they need some improvement.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
I think that using Machine Learning is very good idea and probably the only possible solution at the moment. As I said I don’t know too much about medical usage of computer science in medicine.
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteClassification of viruses into certain groups and finding among groups that are similar to the examined one?
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
As guys mentioned earlier - simulations.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
Maybe DNN. I am not an expert in that field, therefore unable to advise for 100%
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)?
ReplyDeleteThese ways don't have to be technical.
I can not give you answer for this one and that is because I don't have slightest knwoledge about bacteria, so I can not give you proper soluction. But if I have to choose I will go with Deep Learning stuff or even with expert systems like IBM Watson.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
Still I will go with expert systems. https://www.youtube.com/watch?v=hbqDknMc_Bo
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
For me this problem is for Classiffication algorithm (yes I know that this is the most common one) problem so this will be my first choice. And when it comes to software we can go with IBM Watson (again).
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteI believe it is not easy question and my answer is based only on my own speculations. We are not able to track and fight with many kinds of sicknesses. But each known virus is well described and catalogued. The biggest problem is to discover new virus, that can help and control it. As Tomasz wrote - godd idea is AI when we have good dataset
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
As you can see in previous answer when we have proper dataset then best method will be fastest one. I think we have still a big possibilities to find better way that is not known yet ;)
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
I am not sure do we need Machine Learning ... and I repeat. The biggest challenge is to control virus and high quality metrics for this discovery, which gives us valuable input for searching.
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteTo be honest I don't like to be the man who decide on the handling matching those viruses.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
At begining of research I would focus on what software should not to be used for safety reason.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
I like concept of using reinforcement learning to train this program but I fear that this is an utopic procedure. In the beginning machine learning is a good way to attach process and hope the best for future progress. I hardly know answer to this crucial architecture opinion.
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteThe answer is strait forward: simulations. But how to achieve that ? I have no clue.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
No idea. That is not my point of interest. As someone mention above: bioinformatics and other fields of education that focuses on medical usage of computer science. I am not an expert in this field.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
Unfortunately I am not en expert. If you are really interested in that topic please have a look at:
Data Analysis, Machine Learning and Applications: Proceedings of the 31st Annual Conference of the Gesellschaft für Klassifikation e.V., Albert-Ludwigs-Universität Freiburg, March 7-9, 2007
or
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3230886/
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteI can't add anything new here. The only thing that comes to my mind is, as people before mentioned, running computer simulations. With the right data and
algorithms we could with high probability increase the speed of finding matching viruses.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
Once again, like most people before, I have answered this question above.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
Can't say wothout deeper analysis. Probably, due to complicated nature of data some ensemble methods would be unavoidable.
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteWe can use super computers, we have at least a hundred of them across the world. Problem is in data set, do we have enough data to make anything useful?
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
I don't have enough knowledge in this matter, but we can always make something custom for our needs. It will be hard to find something premade which will fulfill all of our needs.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
As I have stated in the previous question, I don't have enough knowledge to answer this question.
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteIt is quite hard to find anything promising. However, I think of simulations which are usually one of the most successful ways of solving problems like these.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
My first thoughts go to the architectures suitable for simulations. I am afraid I can't figure out specific software suitable for activities mentioned in this question.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
I am not a pundit in this field and can't indicate the proper architectures. It depends on the objectives of a particular task.
This is a quite interesting task and surely computer science will play the main role it resolving such questions. It is a future of tackling the problem of drug-resistant bacterias.
I'm not an expert in this scientific discipline. But isn't using viruses and most probably mutating them as well so that they could kill the bacteria a bit dangerous? I mean on the one hand you might find a way to kill a drug-resistant bacteria, but on the other hand, you can cultivate an extrememly dangerous virus that no one and nothing can handle. Of course one can focus on computer simulations, but at some point live expertiments will appear in the picture.
ReplyDelete1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteI think that deep learning and elements related with AI can help to find such a match. Computers with proper algorithms can find such matches similar like they are finding video or audio patterns
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
I am afraid that points 2 and 3 are way over my knowlege
1. Can you think of ways to increase the speed of finding matching viruses (the ones that would kill the bacteria)? These ways don't have to be technical.
ReplyDeleteIn my opinion the fastest way to find out which viruses can be able to kill the bacteria is using of computing cluster to calculate possibilities. Of course it should be supported by deep learning systems.
2. One obvious way would be to use computers to aid finding the right bacteriophage. In your opinion what kind of software would handle this task best?
As I wrote above, deep learning systems can handle this task the best. Nowadays it is possible to use such structures in various situations. Here I see a huge potential for deep learning methods. Moreover software for big data analyse will be needed, too.
3. If your decide on using Machine Learning, what algorithms or architecture you think would be the most promising?
In my point of view either generic algorithms or neural networks can be used however it’s not my cup of tea so I’m not sure… Anyway I still suggest to use deep learning methods.
Thanks for interesting article. Subject looks promising and challenging. It might be a really good idea to support it with good IT solutions. I am not able to propose anything specific, but few quite interesting suggestions were mentioned in previous comments as simulations, supercomputers, or machine learnig.
ReplyDeleteUnfortunately, I do not know on that. From my point of view, the idea is interesting, but difficult to implement and carrying a huge responsibility. I hope that humanity will look new solutions to limit the disease and prevent epidemics. But in such matters must be careful to not generate even more difficult and more serious problem than the one that we want to solve.
ReplyDeleteArticle is very interesting one and gives us new hope on how to fight with diseases using new technologies. Unfortunately I am not a matter export on this topic and I find difficulty on solving problems like that.
ReplyDelete