Projects

A recent work named tldr; introduced a novel scheme to compute the strain enerfy of a molecule, i.e. how much a molecule is relaxed. To achieve this goal it provides, for each chemical bond that can alter the shape of the molecule, the most preffered angles.

The goal of this project is to use these information in the open source molecular docking application muDock. In the current implementation muDock uses a genetic algorithm to dock a molecule inside a protein. In this project the student is expeceted to change how the genetic algorithm behaves by including the information from the state of the art. Part of the experiment is to evaluate the difference in the quality between the current method and the new one.

In our laboratory we are the main mantainers and developers of LiGen, part of the EXSCALATE platform. LiGen is an extreme-scale high-throughput virtual screening application for drug discovery that aims at estimating the interaction strenght between a ligand, i.e. a candidate drug, and a protein that represents the disease. To enable fast prototyping and evaluating new technologies, we also written a small miniapp based on autodock named muDock.

The goal of this project is to port the kernel that implements a genetic algorithm to estimate the 3D displacement of a molecule using Google Highway. Also, the analysis and research of performance portability using other tools like SYCL or HIP can be an option.

In our laboratory we are the main mantainers and developers of LiGen, part of the EXSCALATE platform. LiGen is an extreme-scale high-throughput virtual screening application for drug discovery. This project involves the molecular docking part of the application, which place and evaluate the interaction strength of a ligand (drug candidate) against a target (the disease protein). It has been used for the largest virtual screening up to now. LiGen has been developed in the context of two European Project: Exscalate4COV and LIGATE.

In this repository is available SLM-3dPointCloud. SLM-3dPointCloud is a Python library for generating 3d point cloud holograms, with phase only spatial light modulators, in real time through a GPU implementation. It is of particular interests the Random Superposition algorithm available in the application.

The project consists in the porting of this application using C/C++ from python, and the optimization of CUDA kernel already available in the application. The project involves mainly interaction with high-end NVIDIA GPUs, which we can provide access, along with free courses and certifications from NVIDIA DLI. Also, the analysis and research of performance portability using other tools like SYCL or HIP can be an option.

Accordi Gianmarco
gianmarco.accordi@polimi.it

Palermo Gianluca
gianluca.palermo@polimi.it