Initial draft of singlecell #299
Conversation
The DOI looks okay, may be the double tab that is causing problems
There was a problem hiding this comment.
This is a great first draft. I think we can say more about the deep learning methods referenced here. For example, have these methods been successful? Why or why not? Are there aspects of the neural network architecture that are especially interesting, well-suited for single cell data, or demonstrations of why neural networks are a good fit here in contrast to other machine learning methods?
sections/04_study.md
Outdated
| and even higher-throughput, analyzing hundreds of thousands or millions of cells | ||
| at a time will become reality. | ||
|
|
||
| < still missing a concluding sentence > |
There was a problem hiding this comment.
I suggest converting this to a TODO using the formatting TODO: add concluding sentence. That will help us find it during the next round of review.
I'd also like to add another TODO to look more closely at #153 once we can assess whether it is working well or not. This is already referenced in the gene expression section, but we haven't said much about the success of the approach.
sections/04_study.md
Outdated
|
|
||
| Single-cell methods also promise to uncover a wealth of new biological | ||
| knowledge. Temporality is often difficult to study in biology due to the speed | ||
| at which biological reactions occur. Luckily, a sufficiently large population of |
There was a problem hiding this comment.
Maybe "Fortunately" instead of "Luckily"?
sections/04_study.md
Outdated
| [@tag:Liu2016_sc_transcriptome @tag:Vera2016_sc_analysis]. Finally, examining | ||
| populations of single cells can reveal biologically meaningful subsets of cells | ||
| as well as their underlying gene regulatory networks | ||
| [@tag:Arvaniti2016_rare_subsets @tag:Gaublomme2015_th17]. |
There was a problem hiding this comment.
Whether now or in a TODO, I think we should say more about what the neural network contributes here. Otherwise a lot of the section is an overview of single cell biology and technologies but not directly related to deep learning. There was a lot of discussion in #79 to use as a starting point.
There was a problem hiding this comment.
Sounds good -- will have more discussion in a new draft.
sections/04_study.md
Outdated
| DNA sequence [@tag:Angermueller2016_single_methyl @tag:Qin2017_onehot]. Yet | ||
| another challenge is that these noisy measurements are often made in high | ||
| dimensions, as in transcriptomic or SNP calling studies where there can be a | ||
| thousand cells and ten thousand measurements. Classic dimensionality reduction |
There was a problem hiding this comment.
We cover some of this in the gene expression section as well. Are there differences in the unsupervised strategies we would recommend for for bulk versus single cell data?
There was a problem hiding this comment.
Not really -- a lot of the autoencoder discussion was already in the gene expression section so I removed it.
sections/04_study.md
Outdated
| algorithms could be trained on the evolutionary dynamics of cancer cells or | ||
| bacterial cells undergoing selection pressure and reveal whether patterns of | ||
| adaptation are random or deterministic, allowing us to develop therapeutic | ||
| strategies that forestall resistance. As sequencing and profiling become cheaper |
There was a problem hiding this comment.
Is the point that training datasets for unsupervised and supervised method will grow, leading to more powerful models? That seems harmonious with one of @gwaygenomics's conclusions in the gene expression section.
There was a problem hiding this comment.
I think I wrote something similar to that in the splicing section -- here I wanted to say that single cell data is unique because you have lots of cells, and lots of information for each cell. I tried to make it clearer in this new draft
added more NN discussion
|
@bdo311 Thanks for the revisions. This looks good to me. Are there other changes you wanted to make or can I merge this? |
|
I think this looks good -- fixed the formatting issue. Thanks! |
This build is based on 76857d0. This commit was created by the following Travis CI build and job: https://travis-ci.org/greenelab/deep-review/builds/220505672 https://travis-ci.org/greenelab/deep-review/jobs/220505673 [ci skip] The full commit message that triggered this build is copied below: Initial draft of singlecell (#299) * initial draft of singlecell * Trying to fix tag build error * added more NN discussion * TODO formatting * formatting
This build is based on 76857d0. This commit was created by the following Travis CI build and job: https://travis-ci.org/greenelab/deep-review/builds/220505672 https://travis-ci.org/greenelab/deep-review/jobs/220505673 [ci skip] The full commit message that triggered this build is copied below: Initial draft of singlecell (#299) * initial draft of singlecell * Trying to fix tag build error * added more NN discussion * TODO formatting * formatting
Would love comments for this section as well! @cgreene @agitter