Include feedback and examples from external experts in the lessson

[malvikasharan]

In this repository, we are developing training material for bioscience researchers who are interested in managing and supervising computational projects. We are creating a first draft in February-March 2022 therefore the repository and webpage would look empty. We want to take this chance to invite more opinions, examples and interest to collaborate. 🌸

Especially, we are looking for some help in integrating biological contexts and examples from computational biology and use of data science in life science contexts in our materials. Please reply below to share your thoughts and resources with us.

This project is led by me as a part of our work in The Turing Way and hence, we apply community and acknowledgement practices from The Turing Way, see details.

Everyone who contributes to this repository is acknowledged using 'all contributors bot': see Contributors table in README.

[jcolomb]

I have been trying to implement the use of GIN (a gitlab-similar tool) for RDM in the lab. See https://github.com/tonic-team/tonic.homerepository for the vision (template+git + synchronisation scripts + maybe submodules).

Issues are:

• repo went too big for git, even when using git-annex for file >10MB
• git-annex and windows do not like each other, tool installation on windows is a nightmare.
• we did not have much interactions between the researchers, as only few used the tool.
• all the advantage for coding were not leveraged, because researchers are not coding efficiently (running part of the code, instead of creating a new script)

I am now seriously thinking submodules are needed, and look into using datalad in the background.

PS: There is a 95% chance that I will run a long retreat next year to train people on exactly these subjects, will be happy to contribute here.

[malvikasharan]

Thank you so much @jcolomb. This is fantastic case study and will be helpful for the RDM chapter (and maybe implementation as well). CC-ing @LydiaFrance who has been collating notes from interviews we have had in the last week.

Crossreferencing the summary of interview from the project management repo here for reference: https://github.com/alan-turing-institute/data-training-for-bioscience/issues/23

[malvikasharan]

Add following members as contributors:

1. Victor Tybulewicz:

   • Lab page: Signal transduction in lymphocyte development and activation. Genetics of Down Syndrome
   • https://www.crick.ac.uk/research/labs/victor-tybulewicz
   • References shared: Analysis of congenital heart defects in Down syndrome. This involved analysis of both embryonic hearts from human fetuses with Down syndrome, and from mouse embryos that model the human condition. These were analysed using bulk RNAseq and in some cases using single cell RNAseq. https://pubmed.ncbi.nlm.nih.gov/26765563/,
   • https://elifesciences.org/articles/11614
   • https://www.frontiersin.org/articles/10.3389/fcvm.2021.683074/full

2. Radoslav Enchev:

   • Lab page: https://www.crick.ac.uk/research/labs/radoslav-enchev
   • References shared: Both on neural network
   • Zhong, E. D., Bepler, T., Berger, B., & Davis, J. H. (2021). CryoDRGN: reconstruction of heterogeneous cryo-EM structures using neural networks - Nature Methods. Nature Methods, 18(2), 176–185. doi: 10.1038/s41592-020-01049-4
   • Data availability: Trained cryoDRGN models and generated volumes were deposited in Zenodo at https://doi.org/10.5281/zenodo.435528452. Input files for training (excluding particle stacks) were deposited in Zenodo at https://doi.org/10.5281/zenodo.4412072 and are also available at https://www.github.com/zhonge/cryodrgn_empiar53. We used the following publicly available datasets …
   • Code availability: CryoDRGN software and analysis scripts are implemented in custom software deposited in Zenodo at https://doi.org/10.5281/zenodo.4355743 and are also available at https://www.github.com/zhonge/cryodrgn54.
   • cryoDRGN, an algorithm that leverages the representation power of deep neural networks to directly reconstruct continuous distributions of 3D density maps and map per-particle heterogeneity of single-particle cryo-EM datasets. http://cryodrgn.csail.mit.edu/.
   • Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., ...Hassabis, D. (2021). Highly accurate protein structure prediction with AlphaFold - Nature. Nature, 596(7873), 583–589. doi: 10.1038/s41586-021-03819-2
   • Data availability: https://www.nature.com/articles/s41586-021-03819-2#data-availability
   • Code availability: https://www.nature.com/articles/s41586-021-03819-2#code-availability
   • AlphaFold greatly improves the accuracy of structure prediction by incorporating novel neural network architectures and training procedures based on the evolutionary, physical and geometric constraints of protein structures. https://alphafold.ebi.ac.uk/
   • AlphaFold source code and Colab notebook can be used to predict the structures of proteins not yet in AlphaFold DB. Both resources were recently updated to support multimer prediction.
   • Has lab members working with computational tools, but don’t call themselves computational biologists. Collaborate with the CryoDRGN team to create data that can be combined with tools.
   • Would like to collaborate and contribute to the project.

3. Francesca Ciccarelli:

   • Lab page: cancer genomics https://www.crick.ac.uk/research/labs/francesca-ciccarelli
   • References shared: computational genomics to rebuild cancer evolution; imaging analysis of the TME; ML learning approaches to predict cancer drivers

4. Florencia Iacaruso:

   • Lab page: We aim to identify fundamental principles of how neural circuits are organised and advance our knowledge about the neural basis for guiding attention in a multisensory world. https://www.crick.ac.uk/research/labs/flor-iacaruso
   • References shared: https://github.com/MouseLand/suite2p

5. Evangenline Corcoran

   • Lab: Improving predictions of the impact of climate change on UK agriculture and food security by applying machine learning-based models to image data at various scales (phenotyping for individual plants, satellite imagery for field scale) https://research.qut.edu.au/qase/members/evangeline-corcoran/
   • References shared: Example notebooks for classifying plant patches with MapReader models (https://github.com/Living-with-machines/MapReader/tree/main/examples/classification_plant_phenotype). Demonstrates with a fairly simple use case (classifying patches with plant vs. patches without plant) how a model developed for image analysis in a very different scientific domain (historical maps) can be applied in an agricultural context to extract information that could improve models of plant development.

6. Jim Maas:

   • Lab page: Machine/Deep Learning in plant genonmices
   • References shared: https://developers.google.com/machine-learning/crash-course