if at all my approach of giving better prediction integrating clinical and genomic data is applicable, can some one suggest how to move on with my work?
1.related to dataset collection(whether benchmark datasets are enough or should approach hospitals for claiming it(if yes,how can i ))
2. research gaps that needed to be concentrated and reliable to work.
3.Being CSE student can i defend my work .
Thank you for taking it to note !!
That's definitely possible, for example this past DREAM challenge: https://www.synapse.org/#!Synapse:syn6187098/wiki/
Whether it's reliable or not depends on what question you're asking, and the data/metadata quality. You could cross-validate your results and see whether or not the addition of clinical data (for example) increases accuracy.
1) : I think it's fine to use publicly available data, unless you need a highly specific type of data (e.g. a rare type of cancer).
2) Do a literature search on cancer + machine learning, see what's been done and what can be improved. I would start with this step actually, so that you have a clear idea of what you want to develop.
3) You should show some improvement over the state-of-the-art algorithms/methods. For example, say the problem is to identify which genetic mutation are cancer causing and not, you should have a gold standard data set of known cancer causing mutation and negative controls that you can compare your method to and show an increase in performance at predicting correct class of mutation. If you find papers describing existing ML methods, there should be some validation method that you can probably draw some inspiration from.
this is definetely a need help sir,thanks a lot for very informative insight..!
- Badges
- Science topic