Skip to content

Design and development of Peptide drugs against falciparum Malaria and a Deep learning Web App for Malaria Diagnosis

License

Notifications You must be signed in to change notification settings

igemsoftware2020/IISER-Pune-India

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 

Repository files navigation

iGEM-IISER-Pune-India

Anopheles | The Half-Blood Princess

Project Description

In January 2020, The World Health Organization estimated that 228 million people contracted Malaria globally, 405,000 people died from it in 2018 and accepted evidence for growing drug resistance against Plasmodium falciparum Malaria, especially in South-East Asia.[1] Our project aims to develop a library of inhibitory peptide (known for their high specificity and longer duration against drug resistance) drugs against essential Human-Parasite protein interactions. We intend to use Cyclotides (a type of stable plant proteins) as a protein scaffold for the delivery of these peptides.

  • Using Insilico modeling and simulations, our dry lab team has designed short peptides that will potentially inhibit the protein interactions crucial for the invasion and survival of malaria parasites inside a human host.
  • Our wet lab team has designed various experiments to clone and express the interacting host and parasite proteins, characterize the drug and reduce the toxicity of the grafted cyclotide.
  • To address issues related to poor diagnostics, we have developed a Diagnostic tool using Convolutional Neural Networks and advanced deep learning algorithms to identify patients with malaria based on images of their blood smears.

All scripts, programs and description of software built and used for file processing, analysis, Molecular Dynamics simulation and Deep Learning algorithms can be found on this repository.

Inhibitor Peptide Design

We designed peptide inhibitors against essential falciparum membrane proteins and Human endothelial cell surface receptors using a Structural Biology and Bioinformatics approach. Detailed information on the workflow and questions answered can be found on our Modeling page. All files corresponding to this module of our project will be present in the ./Peptides_against_Malaria/ directory. This is assumed to be the root directory for the documentation in this section. In brief, the workflow involved :

1. Literature Search and Mining the PDB database

We wrote RCSB_static_parser.py, a web parser written in Python that statically mines the PDB database and returns all the required information in the form of a compact .csv file (or as preferred by the user). Although better commercial software is available, our web-parser is efficient, easy to handle, and the output is most suitable for further analysis. We believe that this will significantly help future teams to deal with the data deluge in biology and in particular the RCSB Database.

2. PDB file Preprocessing and Scoring

Under ./1_Preprocessing_and_scoring/ we present the scripts we used to perform preprocessing, computational saturated mutagenesis, scoring of peptide inhibitors and further data analysis. We have written our Python notebooks in a format that is easily readable to new teams and amateurs to help getting started with similar projects.

3. Molecular Dynamics Simulations

Under ./2_MD_simulation_and_analysis/, we have presented the scripts used to perform MD runs and analyze the huge amounts of MD data (~5-10 GB) for each inhibitor model using GROMACS 2019.1 on the Param Brahma facility under the National Supercomputing Mission, Govt of India located at IISER Pune. Our objective was to understand the behavior of the inhibitor over time and determine the specific regions or atoms of the motif that are responsible for the binding.

4. Circularising a Peptide in Cyclotides

We grafted our Peptide Inhibitor in loop 6 of Cycotide kalata B1. The insilico modeling was performed using MODELLER[2] and few custom scripts. We found that such a circularization of a peptide grafted in a cyclotide (in-silico) was never performed. Therefore, we developed our own custom scripts and tested the stability of the drug (cyclotide + peptide inhibitor) by Molecular Dynamics. More specifics on the method followed is on our Grafting subpage of the Wiki. All files are at ./3_cyclization/




DeleMa Detect

DEep LEarning for MAlaria Detection

DeleMa Detect Demo


DeleMa Detect is our Artificial Intelligence based end-to-end Deep Learning software for the diagnosis of Malaria through blood smear images. Our envisioned end-to-end diagnostics system consists of a Foldscope, an easy-to-assemble and very cheap optical microscope, a paper centrifuge that will help in creating the blood smear and this Web Application that uses a Python backend to classify blood smear images taken from any phone as infected or not. Our computational team has developed deep learning software with an accuracy of ~96%. Our goal is to present a web application, that is fast on boot-time and processing without compromising on accuracy so that it could be used on all kinds of mobile phones and in regions with poor internet connectivity as well.

Further documentation and tutorials on how to use it is under ./DeLeMa_Detect. Detailed documentation on how the deep learning model was built by a Transfer Learning algorithm using the Mobilenet_v2 architecture[3,4] is available on the Software section of our wiki.

A miniature version of our Web application is already hosted on Heroku (a cloud-based platform as a service) at delema-detect-igem-iiserpune. Some blood smear images for trial can be found on the Software page of our wiki.

References

  1. World Health Organization: WHO. (2020, January 14). Malaria. The World Health Organisation. https://www.who.int/news-room/fact-sheets/detail/malaria
  2. A. Šali and T. L. Blundell. Comparative protein modelling by satisfaction of spatial restraints. J. Mol. Biol. 234, 779-815, 1993.
  3. Team, K. (2020, May 12). Keras documentation: Transfer learning & fine-tuning. Keras.Io. https://keras.io/guides/transfer_learning/
  4. MobileNetV2: The Next Generation of On-Device Computer Vision Networks. (2018, April 3). Google AI Blog. https://ai.googleblog.com/2018/04/mobilenetv2-next-generation-of-on.html



For clarifications and queries -- Mail iGEM-IISER-Pune@2020