Synthetic Daisies

October 20, 2020

ASAPBio Session on the "Past, Present, and Future of Preprints"

For Open Access Week 2020, Synthetic Daisies will feature an exciting panel discussion on preprints. On Monday (19th), I was part of a panel called "Past, Present, and Future of Preprints", hosted by ASAPBio. I live tweeted the event from the Orthogonal Research and Education Lab Twitter account. If you were not able to attend in person, the recording is on YouTube! The session started with a short introduction from each of our participants: Antonis Rokas, Soumya Swaminathan, Richard Sever, Ross Mounce, and Anjana Badrinarayan.

Yamini Ravichandran and Marco Fumasoni started us off with a short introductory presentation, followed by an introduction by each of our panelists. This part of the session culminated with Marco posing an initial question to the panel.

It turns out that there are many contributing factors to preprint adoption. Some of them involve legacy patterns from manuscript submissions and publications. But preprints also democratizes access to both the production and consumption of scientific literature. It turns out that cultural traditions (within fields and countries), researcher agency, and community incentives are also quite important.

The theme of research culture came up time and time again. But research culture is not only a motivating factor; pro-preprint behaviors can lead to other virtuous practices. For example, Ross Mounce suggests that preprints can encourage a culture of versioning, where different versions of a paper are viewed as important steps in the research process rather than simply being erratum.

There was also a discussion of the role traditional journals play in the research dissemination process. One future direction of preprint culture is to decouple papers from journals. Towards the end of our session, we heard a choice quote from Antonis Rokas and the Rokas Lab.

This combines nicely with observations earlier in the session regarding citation metrics: with the movement towards iteratively-developed preprints with multiple supporting components (open data sets, supplemental figures and notes), there will be a need to distinguish article quality from journal quality. Altmetrics are one path forward, but a more robust system is needed.

Thanks to everyone for participating! Thanks also go to Sarah Stryeck, Jessica Polka, and of course Iratxe Puebla for being a great community manager! Happy Open Access Week!

UPDATE (11/3): A recording of the session is now on YouTube!

October 14, 2020

Hacktoberfest now live!

Welcome to Hacktoberfest! Check out our DevoLearn repositories and our DevoLearn AI resources. Contribute from now until the end of October. Make five commits during the course of October, and Github [1] has something for you!

Want to participate in Hacktoberfest @ DevoWorm? Look at our issue board for Group Meetings, or look at the contribution guidelines for DevoLearn and contribute a Data Science demo.

Select issues on the Group Meetings issue board (DevoWorm) and the Community Board for DevoLearn are also marked with the "Hacktoberfest" label. Once you decide where to contribute, issue a pull request or communicate your interest as a comment in the issue you want to address!

Thanks to Mauyukh Deb and Ujjwal Singh for their administration efforts, and Github users Abtaha, Jainal09, RaviKarri, RudRajit1729, Joel-Hanson, shreyraj2002, Malvi-M, krishnakatyal, and jesparent for their commits so far!

[1] Github offers a T-shirt as incentive for contributing. Offer only applies to labeled repositories (most of the DevoLearn repositories are eligible).

October 8, 2020

Multidimensional Chess Convoluted to a Pretty Picture

At last Monday's DevoWorm Group meeting, I gave a short lecture on ways to interpret multidimensional data using PCA, tSNE, and UMAP. Here are the slides (below) and the YouTube link. The focus here is on Developmental Molecular Biology, but are generally useful, particularly for comparing methods. Here are the slides with a bonus from Leland McInnes, one of the originators of the UMAP technique!

Click to enlarge.

September 30, 2020

OpenWorm Annual Meeting -- DevoWorm Slides

Today we held the OpenWorm Annual Meeting, which is a time for the Board of Directors and Senior Contributors to meet and discuss the latest developments within the foundation (in this case, activities over the past 1.5 years). Overall, a very inspiring meeting! Here are the slides I presented on progress and the latest activities in the DevoWorm group. If anything looks interesting to you, and you would like to contribute, please let me know.

Click to enlarge.

August 24, 2020

Welcome to DevoLearn!

Accelerate data driven developmental biology research with computational learning models

With Google Summer of Code 2020 almost complete, we can debut our latest endeavor: DevoLearn! DevoLearn is a platform that incorporates a computational analysis platform for embryos of different species, with an annotated collection of secondary datasets (DevoZoo) and educational tools.

While the first part (DevoLearn 0.2.0) is brand new, the other two components (species-specific models and DevoZoo) are revamped versions of resources we have created over the course of the past three years. Thanks to Mayukh Deb and Ujjwal Singh for their efforts this Summer, and Vinay Varma, Siddharth Yadav, Asmit Singh, and Bradly Alicea for their past efforts leading up to this point.

Now let's take a look at the the DevoLearn umbrella:

DevoLearn umbrella. Click to enlarge.

The entire project is hosted as a Github project. The DevoLearn software is also hosted on PyPi, and is available as an open-source software package there. The species-specific models (which includes the existing OpenDevoCell resource) are hosted as Herokuapps, while the DevoZoo (access to secondary dat and educational resources) is hosted as a set of Github pages. Aside from the educational component, we encourage people to use DevoLearn for conducting their own Machine Learning and Deep Learning analyses.

Please also feel free to contribute content to DevoLearn! Contributions in the areas of data science tutorials and other types of quantitative analysis are welcome. We are looking for data science tutorials, educational materials that merge ML/DL and biology, and perhaps even novel analytical models. Thanks to Krishna Katyal for pushing a tutorial on Linux command line basics.

Github repository link Click to enlarge.

Over the past few years, the DevoWorm group has been trying to develop new ways of quantifying development. Some of this has been sponsored by Google Summer of Code (embryo cell segmentation) over the past few years, and this summer (thanks to Mayukh) we were able to revisit topics such as worm movement and embryo networks. The DevoLearn software package is a pre-trained model that enables the discovery of meta-features, or features that transcend the typical features of biological image processing.

PyPi project description link Click to enlarge.

Ujjwal's contributions to DevoLearn include a new resource called DevoWormAI and a retooling of the DevoZoo and educational web interface that has slowly been developed over the past few years. Ultimately, this will also include two educational endeavors from 2019: OpenWorm/DevoWorm Curriculum and DevoWormML. These educational initiatives tie together Machine Learning, Developmental Biology, and Complex Systems Theory. Future directions include working towards theoretical models of quantitative embryo data, such as a Laplacian description of the embryo.

DevoWormAI link Click to enlarge.

DevoZoo 2.0 link Click to enlarge.

So explore and make the most out of this resource! Please provide feedback; we would be interested in your proposals for additions or next steps.