How Cloud‑Based HPC Accelerates COVID‑19 Drug Discovery

Virus, Drug Development and High‑Performance Computing

Since the COVID‑19 outbreak, Alibaba Cloud has offered free high‑performance computing (HPC) resources, including SCC supercomputing clusters and CPU/GPU machines, to public research institutions worldwide to support pandemic‑related research.

Recent drug‑discovery projects on Alibaba Cloud’s E‑HPC platform have performed large‑scale molecular simulations, with the cloud supercomputing team providing technical support.

The article focuses on the drug‑screening stage, showing how E‑HPC enables rapid concurrent processing of massive small‑molecule libraries and introduces the Global Health Drug Discovery Institute (GHDDI) open‑sharing platform built on Alibaba Cloud.

Virus and Drug Development

Viruses are nucleic‑acid (DNA or RNA) and protein particles that replicate by hijacking host cellular machinery. Coronaviruses such as SARS‑CoV, MERS‑CoV and COVID‑19 belong to this group and cause severe respiratory and systemic diseases.

Targeting viral proteins, especially proteases like 3CL and PL pro , is a common strategy: small‑molecule ligands bind to the protease active site, altering its structure and inhibiting viral replication.

Drug Development and High‑Performance Computing

Screening millions of candidate ligands by experiment is infeasible; computational docking and molecular dynamics simulations dramatically accelerate the process. For example, a library of 10,000 compounds would require 15,000 hours on a single CPU, highlighting the need for powerful compute platforms and large storage.

Powerful compute clusters

High‑capacity storage for data and results

Efficient high‑throughput screening also requires cluster software that supports multi‑node, multi‑core parallel execution.

E‑HPC High‑Throughput Drug Screening Solution

E‑HPC provides a cloud‑native HPC cluster solution that integrates Alibaba Cloud compute (ECS, EGS, bare‑metal, SCC), networking (VPC, RoCE) and storage (NAS, OSS, CPFS). Users can self‑service create a high‑performance cluster with root access.

The platform offers various instance types (1–104 vCPU, different memory ratios, GPU/FPGA acceleration) and elastic bare‑metal servers that combine VM elasticity with physical‑machine performance.

Example: Parallel Docking with DOCK6

In a serial workflow, each molecule in a mol2 directory is processed by a loop that generates a dock.in file and runs dock6. The serial script looks like:

for molin in mol2/*; do
    molin_name=`basename $molin`
    cp dock.in $molin_name.dock.in
    sed -ie "/^ligand_atom_file/c\ligand_atom_file $molin" $molin_name.dock.in
    sed -ie "/^ligand_outfile_prefix/c\ligand_outfile_prefix $molin_name" $molin_name.dock.in
    dock6 -i $molin_name.dock.in -o $molin_name.dock.out
done

Running this serially is slow. E‑HPC enables high‑throughput tasks that distribute the work across many nodes and cores. The workflow consists of three steps:

Save all molecule filenames to a list file (e.g., ls mol2/* > molin).

Create a task script task.sh that processes a single molecule using the same commands as the serial loop.

Submit the task to E‑HPC with ehpcarr submit -w 96 ./task.sh molin, which launches 96 CPU cores (automatically spanning multiple nodes if needed).

Task status can be queried with ehpcarr status <job_id>, showing which nodes are running which tasks and their current state (DONE, RUNNING, FAILED, INIT).

GHDDI Open‑Sharing Platform

The Global Health Drug Discovery Institute (GHDDI), founded by the Bill & Melinda Gates Foundation, Tsinghua University and Beijing municipal government, builds an open‑sharing platform on Alibaba Cloud. It uses E‑HPC clusters for simulation, provides separate cloud‑supercomputing accounts for partners, and mounts OSS object storage to the cluster for result sharing.

Results are published by mounting OSS to the cluster and serving files via a web server on an ECS instance.

Conclusion

Drug discovery requires massive compute power; E‑HPC offers a cloud‑native, scalable solution that enables rapid high‑throughput screening of small‑molecule libraries, integrates with storage and web services, and supports dynamic task scheduling across multiple nodes.