Pyridostatin

Inspiration

This competition has a combination of chemistry and computer knowledge which are rarely experienced in other competitions so we would like to participate and play with data

What it does

Our key point is just in the data transformation step. We extract features from the metal linker, organic linker and, functional group as describe in the video. Then we applied regression with a deep learning model

How we built it

Clean missing data by using best guest or drop row if not applicable
Clean zero data by using best guest or drop row
Clean wrong type data
Join metal linker data
Join organic linker data
Join functional group data
Feature extraction by correlation heatmap
Create a deep learning model
Split train/test
Train data and save model and weight
Apply model to test set
Tuning model a bit
Submit result

Challenges we ran into

A lot of missing/incorrect data

Accomplishments that we're proud of

Extract feature from the metal linker, organic linker, and functional group
link
- https://openmolecules.org/datawarrior
- http://www.cheminfo.org/Chemistry/Cheminformatics/FormatConverter/index.html

What we learned

We have learnt the MOF theory and also the practical use of MOF in recent studies.
The strategy to calculate the surface area of the structural data by using probe method.