zilliz-bootcamp / chemical_similarity_search

Converting chemical formulae to fingerprints and then compared for similarity in Milvus.

❗❗ This repo will no longer be maintained, please visit https://github.com/milvus-io/bootcamp ❗ ❗

Chemical Structure Similarity Search

Environment

Component	Recommended Configuration
CPU	Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz
Memory	32 GB
OS	Ubuntu 18.04
Software	Milvus 0.10.0 mols-search-webserver 0.7.0 mols-search-webclient 0.3.0

The previous configuration has been tested and this scenario is also supported in Windows.

Data preparation

Data source: ftp://ftp.ncbi.nlm.nih.gov/pubchem/Compound/CURRENT-Full/SDF. The data source contains compressed SDF files. You need to convert these files to SMILES files. We already prepared a SMILE file containing 10,000 chemical structures test_1w.smi. You can use wget to download the file:

$ wget https://raw.githubusercontent.com/milvus-io/bootcamp/0.10.0/solutions/mols_search/smiles-data/test_1w.smi

Deploy

1. Run Milvus Docker

This demo uses Milvus 0.10.0 CPU version. Refer to https://milvus.io/cn/docs/v0.10.0/cpu_milvus_docker.md to learn how to install and run Milvus.

2. Run mols-search-webserver docker

$ docker run -d -v <DATAPATH>:/tmp/data -p 35001:5000 -e "MILVUS_HOST=192.168.1.25" -e "MILVUS_PORT=19530" milvusbootcamp/mols-search-webserver:0.7.0

Refer to the following table for detailed parameter description:

Parameter	Description
-v DATAPATH:/tmp/data	-v specifies directory mapping between the host and the docker image. Please change DATAPATH to the location of test_1w.smi.
-p 35001:5000	-p specifies pot mapping between the host and the image.
-e "MILVUS_HOST=192.168.1.25"	-e specifies the system parameter mapping between the host and the image. Pease update 192.168.1.25 to the IP address of the Milvus docker.
-e "MILVUS_PORT=19530"	Update 19530 to the port of Milvus docker.

3. Run mols-search-webclient docker

$ docker run -d -p 8001:80 -e API_URL=http://192.168.1.25:35001 milvusbootcamp/mols-search-webclient:0.3.0

Note: Please update 192.168.1.25 to the IP address of the Milvus docker.

4. Launch a browser

# Please update IP address and port per your previous configurations
http://192.168.1.25:8001

How to use

• Initial interface

• Load chemical structures
  1. In path/to/your/data, enter the location of the smi file. For example, /tmp/data/test_1w.smi.
  2. Click + to load.
  3. You can see the number of chemical structures have changed: 10000 Molecular Formula in this set

• Search chemical structures
  1. Enter the chemical structure to search, such as Cc1ccc(cc1)S(=O)(=O)N, and press <ENTER>.
  2. Set the value of topk. This demo returns topk most similar chemical structures.

• Clear chemical structure data

  Click CLEAR ALL to remove all chemical structure data.

Conclusion

This demo implements chemical structure similarity search with Milvus. You can also use your own SMILES data. For 37 million chemical structures/90 million structures (512-dimensional feature vector), Milvus has the following performance:

	Performance (37 million)	Performance (90 million)
Search a single chemical structure	190 ms	480 ms
Search 50 chemical structures. The performance is the average time of each structure.	6 ms	12 ms

The following table shows memory usage:

	37 million	90 million
Memory Usage	2.6 G	6 G

We can see that Milvus has faster search speed and lower memory usage for massive-scale chemical structure search.

We have built a chemical structure similarity search system (http://40.73.24.85) based on Milvus. Welcome to search for your own chemical structure.