Disqus - Latest Comments for pedrobeltrao

Re: Evolution of Cellular Networks: 20 years of open science or how we haven't radically changed the way we do science online

pedrobeltrao — Thu, 17 Nov 2022 01:17:52 -0000

I don't think it is a shallow take, it is probably the biggest reason. The incentives are aligned to the traditional group structure with a PI and lead author from the group. This makes spontaneous collaborations across groups harder to reward even if these days we have an easier time knowing what everyone else could do. The two cases I mentioned were cases when a larger than usual group of scientists got motivated to work on the same topic at the same time and it was easier to get people together.

Re: bioRxiv viewer

pedrobeltrao — Fri, 03 Apr 2020 08:14:32 -0000

Could you please share the list of interactions as a supplmentary file as well ? or provide a link in the comments section to a file containing this information?

Re: Public Rambling: Why would you publish in Scientific Reports ?

pedrobeltrao — Thu, 27 Oct 2016 04:13:20 -0000

I completely understand and the IF trend for PLOS One and Sci Reports is probably the reason why a lot of people are switching from PLOS One to Sci Reports. Nature/Springer journals funnel their rejections from journals that have higher rejection rate (Nature, Nature X) to journals that care less about "impact" which gives them a big advantage over PLOS. As an scientists I also try to publish in journals that I hope will further my career and the career of the people in my group. I wished as well that the journal where we publish was not important but unfortunately it is not the case. I think the trend right now is to move away from IF and to an evaluation based on single articles but we are still very far from having that.

As for the costs, they can be much cheaper than they are today, in particular due to the fact the system is not a free market and brand contributes a lot to the profit. New journals that may have a lower overall cost structure may have a hard time breaking into the market.

Re: Evolution of Cellular Networks: Phylogenetic history of fungal protein phosphorylation – the anti-press release

pedrobeltrao — Fri, 14 Oct 2016 04:53:17 -0000

I agree and that is exactly the ongoing discussion. How many of these poorly conserved phosphosites are really doing anything at all. It is impossible to prove a lack of function so the way we are approaching this is to come up with ways to rank phosphosites according to their potential to have a function. Conservation is a way but there are other ways such as trying to propose mechanisms. Can the phosphosites regulate an interface, a localization signal, are they regulated under certain conditions, etc.

We don't think that the kinase target specificities change that much. A human kinase can often replace the corresponding yeast kinase. What we think is happening is that kinases will have a central group of target sites that need to be regulated and then some other sites that arose by point mutations that get also regulated even if they have no functional role. There is also bound to be some neutral changes via redundant intermediates whereby a kinase switches target sites and still controls the process. This happened for the cell cycle where Cdk1 controls the process in different ways across different species.

Re: Evolution of Cellular Networks: Replace journals with recommendation engines

pedrobeltrao — Tue, 05 Jan 2016 09:40:05 -0000

I agree with you that we could benefit from publishing less. However that is not the trend. If anything, as science as moved online the trend has been to "publish" more and more science with some people arguing that we should micro-publish all of the individual experimental observations (see for example www.sciencematters.io). I don't see a reasonable way to incentivise scientists to publish less so the solution might really have to be better filters instead of fewer publications.

Re: Is "Scientific Data" ever-finer salami-slicing, or is it reducing time to data publication?

pedrobeltrao — Sun, 20 Oct 2013 16:25:02 -0000

I completely understand the value of publishing first the data to allow early access to everyone. I do say that in the blog post :) and I have written about this idea of publishing smaller tokens/bits/units of science many years ago (http://www.evocellnet.com/2... . You are assuming that most people will want to do this. I think the majority of people will still want to hold on to the data until they analyze it and will essentially use this journal to get a second publication for an extra fee. Essentially getting two publications out of the same dataset with waste of reviewing time and funding money.

If most people did the Right Thing and used this journal to publish data for early access while they work on the analysis then there would be value it this. This relies on people wanting to do the right thing since there is nothing stopping them from holding on to the data until they can submit both papers at the same time. However, if everyone did the right thing then we would not need this journal would we ? Everyone would make the data available in the very same repositories that this journal is recommending. Maybe it would work if this journal somehow guaranteed that these datasets are not already being used in an accepted/submitted publication somewhere else. From my point of view, this does not change any incentive structures it just wastes money.

Besides this major point it strikes me that this journal is adding no value whatsoever. Even figshare does more with the data then what they are going to do here. On the flip side, the data infrastructures that this journal will rely upon (community databases) are constantly having problems finding stable sources of funding. So this channel of funding that would be essentially a fee to submit an entry to a database would be a very interesting business model for these databases. If these data journals are to exist then the money should be at least solving a problem of data infrastructure.

Re: Evolution of Cellular Networks: Doing away with scientific journals

pedrobeltrao — Sun, 09 Jun 2013 15:52:07 -0000

I am not sure filtering by power of the study would work. That depends so much on the area. Sometimes a good idea could have a small sample size and still have an impact. I would be more inclined in developing recommendation engines that combine some topic clustering methods, machine learning recommendations and article level metrics to surface content to readers. As readers we could select the number of topics (from a set of recommendations) to read from and depth in each topic. The article level metrics could then be used to surface articles more or less stringently based on the depth we picked.

Re: Evolution of Cellular Networks: Doing away with scientific journals

pedrobeltrao — Sun, 09 Jun 2013 15:47:35 -0000

That is a shame. I always use the physics community and arxiv as the prime example of how a preprint culture can work. I hope we can get other communities to start posting to preprints instead of seeing physics getting out.

Re: Evolution of Cellular Networks: Doing away with scientific journals

pedrobeltrao — Sun, 09 Jun 2013 15:45:15 -0000

I don't know what you are trying to say. Soon after that I say that my opinion could be biased and wrong. I then go on to show that in fact high impact journals are highly enriched in future highly cited papers. So the data supports my subjective opinion.

Re: Public Rambling: Why would you publish in Scientific Reports ?

pedrobeltrao — Sat, 06 Apr 2013 06:59:58 -0000

I don't think the impact factor is going to be that much higher than PLOS One. Nature also has Nature Communications that is already positioned below all other Nature branded journals and that has an impact factor of 7. So Scientific Reports is likely to be getting somewhere below Nature Communications. Nature is trying to reproduce the PLOS system of having "lower tier" bigger journals cross subsidizing the journals with higher rejection rates. In this case they are just taking the best economic model from both worlds. A high rejection rate set of journals that is closed access and every library needs to pay for, and a high volume open access set of journals that bring in a lot of cash per paper published.

Re: Evolution of Cellular Networks: The glacial pace of innovation in scientific publishing

pedrobeltrao — Thu, 28 Mar 2013 12:57:06 -0000

You are right that "discussion" might be a bit too generous but there are some. For example, for the encode papers there was a significant backlash and arguments being discussed on twitter. The meat of the discussion was going on in blogs but I think that Twitter promoted the discussion and gave visibility to some of the more interesting posts. Even now with these Nature articles on the future of publishing you can see how several people started blogging about the topic and sharing related links. So I think it can promote discussion even if the 140 character limit makes it ridiculously hard to express an argument.

Re: Evolution of Cellular Networks: The glacial pace of innovation in scientific publishing

pedrobeltrao — Thu, 28 Mar 2013 08:56:34 -0000

As you say, we don't want to do more work to take better advantage of the web. This is for me the central problem. There are no incentives on the part of authors or publishers. Chance has been happening mostly because of economic reasons like founders mandating open access. Even when publishers try to innovate, like for example the creation of specific social networks for scientists, there is no real usage. I think the next incentive that might work be the need for filters to parse through the PLOS One clones. I would pay for a good recommendation engine even if Google Scholar already has a reasonably good one for free.

Re: Public Rambling: My Californian Life

pedrobeltrao — Thu, 27 Dec 2012 08:18:27 -0000

for those that are good in programming and (statistics or design) have no problems in finding a job in the bay area. For science, there are plenty of jobs but it is has hard or harder than in many other places.

Re: Public Rambling: The Minimal Publishable Unit

pedrobeltrao — Wed, 09 May 2012 16:26:46 -0000

Hi Dave. I guess the point was not so much to "publish" the result as to discuss how much content should go into a paper. Maybe I should have expanded the discussion into how blogs and other publishing platforms like figshare are facilitating self-publishing and indexability.

Re: Public Rambling: Academic value, jobs and PLoS ONE's mission

pedrobeltrao — Mon, 27 Feb 2012 14:25:06 -0000

I totally agree with your last idea and if I had more of a background on text mining and machine learning I would even bet resources on this. I would just add that there are labs that have expertise in these things and I think there is a lot of room for collaborations between publishers and academia on this. What I am a bit surprise is that there is little commercial interest in this. I would pay for a personalized sorting system and I think a lot of people would as well.

Re: Public Rambling: Academic value, jobs and PLoS ONE's mission

pedrobeltrao — Mon, 27 Feb 2012 14:21:12 -0000

I meant primarily the second aspect of focusing our attention on what papers to read based on the journals. I don't want to go into the grey literature plus the potential for open science material from ongoing research. Even if only 30% of articles are currently getting rejected, the fact is that the total number of articles published is growing at an increasing pace.
In any case, to focus on the filtering based on the journals. Let's take for example systems biology as something that I am interested in. It is almost impossible for me to program a search query in pubmed that would give me a clean feed of papers that fit my interest in systems biology but the Molecular Systems Biology journal consistently puts out papers that are almost all interesting to me. The same is not true for example from BMC Systems Biology. It is not even necessarily about impact factors. For example PLoS Computational Biology is much more consistent in putting out papers that interest me over Bioinformatics or BMC bioinformatics. So, in a very consistent way journals like Molecular Systems Biology, PLoS Comp Bio, Genome Research and others put out papers that I find interesting. I would be very annoyed if I had to fish those papers out from a collection of unsorted pubmed entries. As I said before, I try a bunch of different things to find papers of interest that includes pubmed queries and social networks and from my experience nothing works better than some editors/journals that I follow. Some of my social network contacts do a very good job (some are editors) but how much do they filter based on the journals that they choose to follow ?I guess I should also make a stronger separation between keeping up-to-date with the literature and searching. You mention searching a few times and maybe you are talking about finding papers when you are actively researching a direction. This a very different activity from keeping up to date with the literature. When researching in one specific direction you might try to find a review and follow the citations in that review or use citation networks in G Scholar to find the major findings in a field etc. Even in this regard, if you think about how reviews get written you still need to be aware of the ongoing research in a field. So, someone needs to be keeping up to date with the literature and from my perspective this would be impossible with 3000 unsorted entries in pubmed per day. I think I am deviating from the main point but I have heard several people claiming that they don't need journals to find interesting articles but this is so far away from my reality that I feel must be missing something. I just can't see how we would find out about interesting work if all we had was pubmed. Don't forget that your social network would also just have pubmed. I don't see it.

Regarding PLoS ONE's responsibility in filtering, we probably just have to agree to disagree here. As I said before, it was their founding mission to do post-publication peer review. PLoS ONE published 755 articles last month and the only way right now that they let me sort through them is tags. If the idea of PLoS ONE continues to grow we might get a world of a handful of big repositories of unsorted articles and I think they have a responsibility to develop these sorting tools .. as they said they would. I totally agree that they are pushing others in the right direction (open access, re-use, article metrics), they just should be setting an example in this as well.

If we are to replace the current system of tiered journals with one of ranked articles we have to have an article ranking system in place.

Re: Public Rambling: Academic value, jobs and PLoS ONE's mission

pedrobeltrao — Sat, 25 Feb 2012 16:48:56 -0000

I would like to believe this is not an big issue but all the trends point to continuous increase in number of publications. I just check and arXiv gets ~6000 (and growing) submissions per month so that sounds like a big challenge to me already. Besides, it is not like we cannot create these filtering tools. Very advanced versions of these are being used for all sorts of problems (amazon, netflix, google, etc) so why should we not have better solutions ourselves?

Re: Public Rambling: Academic value, jobs and PLoS ONE's mission

pedrobeltrao — Sat, 25 Feb 2012 16:39:31 -0000

Regarding your second point and as I replied to Ross' comment, I think you underestimate how much the journals are used for filtering. If there was one single stream of abstracts I am very sure we would all be in very serious trouble in regards to communicating with each other through papers. Honestly I have to challenge your assumption that you don't need journal filtering for the way you consume scientific news. I explore more than most people I know all sorts of different ways of finding papers to read and I read more than most people I know around me. I can use G Scholar to find older papers in a topic by going through citation networks and I can use pubmed to get RSS feeds on keywords to track some fields that are very close to what I am doing. Even for these fields that are very close to what I do, Pubmed keyword matching is just a horrible simplification that does not work well. As for suggestions I get from Twitter and G+ (unfortunately no longer from GReader contacts), they depend on the journals, only indirectly via my contacts.

Regarding PLoS I agree that PLoS is no the whole literature but the creation of PloS ONE gives them a huge responsibility in creating filtering tools. As you well know, PLoS ONE was created with the mission of establishing post-publication peer review systems. The works would be published with no regard for "impact" and after publication the action of the readers would help sort out the most interesting research. The growth of PLoS ONE caught everyone by surprise and most publishers created clones. This demonstration of an economically viable open access business model is contribution to the attention crash. So I do think PLoS as a responsibility to build the filtering tools. As I say in the post, the very least they can do is change their marketing message. I am sure you have also seen many presentation by PLoS ONE members where they keep pushing this message of how big PLoS ONE is getting, what fraction of Pubmed they are going to be next year. They have to re-focus, I don't care that PLoS ONE is this growing, huge, catch all journal, I want to know how they are going to make it easier for me to figure out what is the most exciting research they are publishing (as per their mission statement).

More generally, I guess I just hope that the filtering problem come up more for discussion. We are so behind many other markets on this (amazon, goolgle, pandora, last.fm, etc) that it is almost a shock.

Re: Public Rambling: Academic value, jobs and PLoS ONE's mission

pedrobeltrao — Sat, 25 Feb 2012 16:15:48 -0000

I completely agree that it *should* be possible to filter based on keywords, machine learning methods trained on previous papers, social media, etc. My point is that we have to build these systems up if we want to do away with the journal tiered system.It *should* be possible but we hare not even close to having this. You should also not under-estimate how much your social contacts also depend on the journals for filtering.

I also use twitter, G+, RSS feeds from Pubmed on keywords .. I even have a Yahoo Pipe doing pattern matching on a subset of journals. Overall it still basically sucks and I think it would all still collapse if there were no journals, if there was just one big stream of abstracts.

Re: http://blog.readcube.com/post/12478969191

pedrobeltrao — Wed, 16 Nov 2011 13:23:48 -0000

Please let us know when a linux version comes out. thanks

Re: Public Rambling: Nature Communications serves its first papers

pedrobeltrao — Fri, 06 May 2011 14:15:20 -0000

NPG did not have a large stable of open access, author-pays, journals. That is also reflected in their more recent decision to create the journal Scientific Reports, an PLoS ONE clone. I think their decision to create Scientific Reports further supports my arguments here.

Re: Data matters but do data sets?

pedrobeltrao — Fri, 01 Apr 2011 18:50:46 -0000

I am not sure I get it. Getting datasets and mashing them together to answer my questions is essentially what I do all the time. An example ... lets say I want to see the impact of protein abundance on the evolution of protein interactions .. I would take some dataset that was published on protein abundance, two datasets for protein interactions in different species and put it all together to look at my problem. The datasets were not created to answer this question but they are available for any idea that I might have. I can't even cite individual protein abundances , there aer no IDs for that. One other very common example is making predictors by combining different datasets .. lets say I want to predict protein-protein functional interactions I would gather things like protein localization, expression data, physical interaction data from different methods and use some machine learning approach to weight the different data types. All of these are different datasets that were publish to answer some specific question or essentially as a resource for others to use. I think datasets are of very valuable exactly because they create a digital pool of observations that others can use. This is why dataset papers are so widely cited as well. Maybe I missed your point completely ?

Re: Public Rambling: Structure based prediction of kinase interactions

pedrobeltrao — Tue, 08 Mar 2011 14:14:30 -0000

Hi Chris,
From what I understand I think the method still requires some knowledge of domain-peptide interactions to derive their predictions but they try to predict what residues in which positions of the kinase domain might drive the specificity. They don't predict all kinases so I assume they have some way to tell which ones they don't have enough information to make the predictions. I wish I had more time to re-implement their approach and play around with some ideas. This would be cool to look at the evolution of kinase specificity.

Dave Bridge: Have a look at Networkin's paper (http://networkin.info/) for a good description of the current best guess. Essentially you would guess that the specificity of kinases in vivo depends on the substrate recognition plus things like co-localization, co-expression etc. In fact, taking these things (or some approximation of them) into account really improves the in vivo kinase-substrate predictions. Mike Jaffe had a really cool example at a recent meeting on the specificity of mitotic kinases. I think it is not yet publishes but essentially he showed that kinases that had similar localizations had non-overlapping substrate preferences but those kinases that did not share cellular localization had more similar binding preferences. Having said all of this it could still very well be that the current experimental methods are not yet giving us the complete picture on the substrate recognition, maybe the different kinases have very peculiar secondary contacts, etc.

Re: Public Rambling: The GABBA program

pedrobeltrao — Wed, 22 Dec 2010 19:38:05 -0000

Hi Martin. I don't think the program organizers are really keeping track. From my generally impression I would say that most people end up doing the postdocs abroad as well. I know of several examples (from the GABBA but also from other similar programs) that have returned to Portugal as group leaders. I suspect that currently the limitation is mostly on the lack of positions to receive those getting trained abroad. The positive side of this is that there is a talented pool of alumni to tap whenever positions open.

Re: Deriving Inhibitory Peptides from Globular Protein–Protein Interactions

pedrobeltrao — Wed, 06 Oct 2010 16:40:40 -0000

Nice work. I did not read the paper and maybe you talk about this it in but I have a question regarding these linear segments. Do you know of cases where a domain interacts with both a linear segment within a folded domain and a similar linear peptide within disorder regions ? In other words, do you think these domains are also capable of forming domain-peptide interactions in a way we are used to thinking about for SH3/PDZ/SH2 domains etc ?