Review summary

The proposal lacks a cohesive flow. I don't know how all the pieces (Bcr-Abl, CML, crowdsourcing, machine learning, NLP) fit together. I recommend telling a story. Here's an off-the-cuff example:

• CML is a big problem.
• The solution lies in Bcr-Abl.
• Lot's of research has been done about Bcr-Abl, so much that no one expert can know it all.
• In fact you couldn't even get many experts to write down all the knowledge on Brc-Abl.
• The easiest/cheapest/fastest/most-scalable method for retrieving all the knowledge on Brc-Abl is to use automated text mining.
• However automated text mining alone is unreliable.
• Crowdsourcing (getting non-experts who are cheap to read through literature) allows us to evaluate and improve on the NLP. Crowdsourcing provides the necessary feedback to learn how to do the automated text mining.
• The literature mined information fits into a network.
• There is a network algorithm that uses the network topology to predict the solution to CML.

You need to present each stage as the most logical next step. Additionally, it would be nice to address the shortcomings of alternatives. For example, what is deficient about current text mining networks relating to Bcr-Abl? Shouldn't you try those first? Preferably, you can answer these questions about alternatives as part of the story.

Best of luck — I think the proposal combines a bunch of cool and cutting edge techniques. Now you just need to communicate the logic of the approach in a clear and accessible way.

Hi Daniel,

Thanks for the helpful review. I agree that the proposal lacked cohesiveness and am working on a revised version now.

I agree that the proposal lacked cohesiveness and am working on a revised version now.

Awesome, can't wait to see the next draft. You're the first person to receive in depth proposal review on Thinklab, so I'm excited to see how helpful you find it. I can imagine so much feedback (currently four detailed reviews) can be a bit overwhelming, but I suspect the proposal will strengthen as a result. Consider posting your experience when done: what worked, what didn't work, and whether you'd recommend public proposal review to others.

One further piece of feedback: the techniques you propose implementing, specifically crowdsourcing and machine learning, have high upfront costs. Therefore, I feel that it's especially important to justify that they're essential to solving the problem. Otherwise, the grant reviewers may hesitate. Instead you want the reviewers to "hesitate not, fund or fund not".

I would be more than happy to write up my thoughts about the review process and Thinklab itself after I submit the proposal. I will also be providing a reference to these discussions in the application itself, so maybe the HHMI reviewers will visit the site.

As for the proposal itself, I will work to have a revised version posted before the application deadline, but don't know if I will be successful. I've decided to alter the direction slightly to go after why CML stem cells are not killed by Bcr-Abl inhibitors, since after a further review of the literature it seems that this is where the field is going.

I will also be providing a reference to these discussions in the application itself, so maybe the HHMI reviewers will visit the site.

Great idea. I think it will impress the reviewers that the proposal has already been publicly vetted by peers.

Best of luck with the new direction!