Library article / world-community-grid

World Community Grid — Citizen Computing for Science

Trace WCG from its 2004 IBM launch and transfer to Krembil Research Institute through BOINC work distribution, validation, medical and climate research, results, and current operations.

12 min readLast updated: August 22, 2026Last verified: August 22, 2026

Key points

World Community Grid (WCG) directs spare computing capacity donated by people around the world toward medical, environmental, and other research. Behind its enormous cumulative compute totals is a longer chain: selecting projects, dispatching work to heterogeneous hosts, validating suspect results, downstream analysis by researchers, and public release.

01Not one experiment, but shared infrastructure for public-interest research

World Community Grid is not the name of one giant science experiment. It is shared infrastructure that, depending on each project's support, sends research applications to one or more of Windows, macOS, Linux, and Android devices, then returns outputs to research teams. The client uses BOINC, while WCG runs project selection, servers, application distribution, statistics, and community operations.
In the official project statistics updated at 12:06 UTC on 21 August 2026, three research projects are active: Mapping Arthritis Markers, Africa Rainfall Project, and Mapping Cancer Markers. Twenty-nine are completed and none is shown as paused. Beta Testing is intermittent, but it is a cross-project proving ground for applications and workunits rather than a separate research study.
At the same snapshot, the cumulative display reports 821,691 registered members, 7,747,340 registered devices, 2,732,527 years of runtime, and 7,897,462,759 returned result instances. These are totals across more than two decades, not counts of people or devices currently active. Runtime and result count measure compute contribution; they do not count treatments or scientific conclusions.

02Grid, BOINC, and WCG are not synonyms

Layer	What it means	Its form in WCG
Grid computing	Coordinating multiple computing resources toward a shared purpose	Aggregating heterogeneous, intermittent hosts around the world for research
Volunteer computing	People choose to donate their own resources	Participants provide power, devices, and connectivity
BOINC	Middleware for dispatch, execution, and return of work	Foundation for client and server functions
World Community Grid	A research portfolio and operating community	Proposal review, application operations, validation, statistics, and links to research teams

WCG is not a network without central administration. WCG selects studies, distributes signed applications, schedules workunits, and judges returned results through validators. Its hosts are geographically distributed, but authority and coordination are explicit. That differs both from an institutional grid, where organizations federate resources through contracts and identity, and from Bitcoin, where open participants compete to order a ledger.
WCG's achievement is therefore not eliminating the center. It connects a large population of diverse, unpredictable hosts to accountable science projects under safeguards such as review, signing, and validation as documented by the operator. That is not a third-party audit or a guarantee of zero risk. Distribution is not binary: ownership of resources, choice of work, validation, and publication of results must be assessed as separate layers.

03From the Smallpox Grid to IBM infrastructure at world scale

WCG moved from its 2004 IBM launch through BOINC integration, legal transfer to Krembil in 2021, and an operational handoff in 2022. Separating ownership from system migration is essential to reading its continuity accurately.

The direct predecessor was the Smallpox Research Grid on United Devices' grid.org. An Oxford University team used computational chemistry to search for compounds that might bind to smallpox-virus targets. The scale of that participatory effort helped motivate a continuing grid for humanitarian research.
WCG launched on 16 November 2004 as a philanthropic initiative of IBM Corporate Social Responsibility. It initially used United Devices technology, added Linux and BOINC integration in 2005, added Windows support on BOINC in 2006, and later supported macOS. The United Devices client was retired in the third quarter of 2008, leaving BOINC as the common platform.
This history shows middleware born from SETI@home moving into a separate institution and providing one participation interface across medicine, the environment, and basic science. BOINC did not thereby decide each project's scientific application or validation rules; the research team and WCG operators retained those design duties.

04Choose the research before gathering compute

WCG proposal criteria ask for more than a large workload. A study should benefit humanity, be led by a nonprofit research organization, divide into many independent computations, and make results obtained through WCG available to the scientific community. The value judgment — what spare CPUs should do — therefore happens at the entrance to the technical pipeline.
Acceptance does not mean arbitrary code is immediately dispatched. Researchers and the technical team port and test the application, then design inputs, outputs, checkpoints, error handling, validation, and server-side ingestion. Beta Testing exists as a distinct pool because household operating systems, CPUs, drivers, and runtimes are far more varied than a managed cluster.
Even with a publication policy, raw returns do not instantly become scientific knowledge. Opening results directly generated through WCG to the scientific community is not the same as unconditionally releasing input clinical raw data, every line of source code, or every downstream artifact. Researchers reconstruct results, check quality, perform statistical or experimental validation, and publish data or papers. End of computation, WCG project completion, peer-reviewed publication, and clinical use are separate milestones.

05How one workunit becomes research data

WCG does not stop at dispatching workunits. A research question is split, applications are reviewed and signed, volunteer hosts compute, outputs are validated or retried, and researchers aggregate, analyze, test where needed, and publish. Completed computation and scientific conclusion are different stages.

A research team splits a large search space or simulation into smaller tasks and makes them independently schedulable where dependencies permit. In ARP, for example, one generation can feed the next. WCG scheduling considers platform, resources, deadlines, and project choices before dispatching tasks. The BOINC client downloads the application and input, runs in the background, and checkpoints when supported.
Finishing a computation and uploading an output file is not the end. The client reports it to the scheduler and WCG validates it. Validated output passes through project-specific assimilation and post-processing before reaching researchers. Where aggregation happens and how results are reconstructed vary by project.
Unresponsive hosts, missed deadlines, hardware faults, and calculation errors are normal operating conditions. Work must be reissuable, while central servers face distinct bottlenecks in generation, dispatch, upload, databases, validation, and assimilation. Adding hosts does not guarantee proportional research speedup.

06Replication, trusted devices, and signatures divide trust

WCG validation is not one universal double-computation rule. Redundant copies and single validation on trusted devices with random replication audits feed application-specific validators. Either method can produce a valid result or trigger extra copies and retries after mismatches, deadlines, or numerical differences.

WCG help does not prescribe one validation method; the figure shows two major paths. Redundant computation normally sends a job to two devices, accepting agreement and issuing more copies on disagreement. Single Validation Type 1 sends one copy to a device with a reliable record while random replication audits its trust state. Type 2 is also defined for nondeterministic output using multiple copies, though the official help says no current project uses it.
This validation protects consistency of returned computation. It does not establish the research hypothesis, remove bias in the input dataset, validate a statistical method, or demonstrate clinical utility. Those questions belong to study design, downstream analysis, replication, peer review, and where necessary experiments or clinical research. Validated credit is not a vote on scientific truth.
Trust also runs in the other direction: participants execute WCG-distributed code on their own devices. WCG documents application review, cryptographic signing, rejection of unsigned applications, and encrypted communications. A signature helps show that a binary matches what WCG distributed; it does not automatically prove research value or absence of bugs. Participants still need to inspect the operator, research page, and security notices.

07Three active projects in August 2026

Official statistics at 12:06 UTC on 21 August 2026 show three active projects, one intermittent Beta Testing pool, 29 completed entries, and none listed as paused. The active work covers arthritis markers, African rainfall simulation, and cancer marker combinations; every status needs its observation time.

Project	Question	Role of computation	Official status (checked 2026-08-22)
Mapping Arthritis Markers	Which markers in psoriasis and psoriatic-arthritis skin samples may relate to PsA development, progression, or treatment?	Search candidate patterns in skin-sample data	Just Launched / Active
Africa Rainfall Project	How can localized rain over sub-Saharan Africa be represented at high resolution?	Run many high-resolution, regional weather-simulation workunits	In Progress / Active
Mapping Cancer Markers	Which marker combinations relate to cancer type, outcome, or treatment response?	Search and score combinations at large scale	In Progress / Active

This is the portfolio classified as Active in the current official registry, not a ranking of scientific value. Active does not guarantee a constant queue for every platform: research batches, server maintenance, input generation, storage, and downstream analysis all change availability. “My client received no task” and “the project has ended” are not equivalent.
A WCG project name also need not cover the whole scientific program. WCG supplies one computational stage; sampling, observation, dataset preparation, experiments, statistics, and publication sit before and after it. Volunteer computation is an important stage, not a substitute for the rest of research.

08From MCM to MAM — distributing combinatorial search

Mapping Cancer Markers (MCM) searches DNA, RNA, protein, and related marker data from healthy and cancer tissue samples for combinations that may correlate with cancer type, outcome, or treatment response. The number of combinations explodes as markers are added, so mathematical methods narrow the space while distributed jobs score many candidate signatures.
Its output is not itself a diagnostic or therapy. It is a set of candidates and search methods for later validation. A June 2026 official update says lung-cancer analysis yielded 26 high-scoring genes; a high score is not clinical approval. Establishing clinical meaning would require additional tests such as independent cohorts and biological validation.
Mapping Arthritis Markers (MAM) applies the large-scale marker-search method developed in MCM to skin samples from people with psoriasis and psoriatic arthritis, looking for patterns that may relate to PsA development, progression, or treatment. What the official record supports as inherited is the computational method and algorithm — not transfer of individual medical candidates between diseases.

09Africa Rainfall Project — storage can bottleneck before compute

Africa Rainfall Project (ARP) runs high-resolution weather simulations of localized rain over sub-Saharan Africa and compares them with satellite, ground-observation, and The Weather Company data. Its long-term aim is to improve models that can support understanding of agriculture and water resources where rainfall observations are sparse.
The project made visible that volunteer CPU is not the only limiter. Storage on the research side reached capacity for its large outputs, and new work stopped in December 2022. An October 2024 notice targeted restart by 4 November, and a July 2025 official update confirmed that the project was running again.
Even when edge computation seems nearly free, central networking, databases, storage, backup, and researcher time remain. ARP's pause is less a story of failure than an operational example of the pipeline bottleneck moving from host computation to aggregation, storage, and post-processing.

10“Completed” does not mean a finished medicine

Completed in the official statistics means the stage of continuing to distribute new WCG workunits has ended. Organizing returned data, follow-up experiments, writing, and replication can continue. A search that finds no expected candidate may still have scientific value if it publishes its explored space and method.
Peer-reviewed outputs still require a narrow statement of what they established. FightAIDS@Home reported a binding-free-energy method using about one million replicas on WCG, not an approved HIV therapy. Help Conquer Cancer used WCG to compute 12,375 primary features per protein-crystallization image and evaluated a classifier built from them on a held-out set, while its authors warned that precision would fall in deployment because actual crystal images are rare. OpenZika narrowed more than 400 million computation results into candidates tested in enzyme and cell assays, but remained before hit-to-lead development.
Evaluation should separate the initial question, computation performed by WCG, validation, released data, peer-reviewed papers, independent replication, and distance to practical use — rather than asking only how many compute-years were donated. A candidate compound is not a medicine, and a computational association is neither causation nor a diagnostic. Keeping those boundaries is how volunteer contributions can be recorded honestly rather than inflated.

11From IBM to Krembil — continuity required an operational transfer

IBM announced in 2021 that WCG assets would transfer to Krembil Research Institute, part of Toronto's University Health Network (UHN); the legal transfer completed that September. Operations moved to an academic team centered on Igor Jurisica's lab. IBM funding continued through February 2022, and extended transition support ended on 2 March.
For the move, new work stopped on 14 February 2022, result receipt on 27 February, and the website entered relocation on 28 February. The website and forum returned on 21 June, and a complete restart after alpha operations was announced on 13 April 2023. Database, scheduler, storage, and performance problems exposed the difficulty of moving a corporate-scale service onto research-institution resources.
The current security page says physical servers are hosted in the SHARCNET facility at the University of Waterloo and the technical team is based in the Jurisica Lab at Krembil Research Institute. Infrastructure descriptions can change, so prospective participants should use current news and server status. WCG survived two decades through more than client code: a receiving institution, technical staff, researchers, and a forum community all mattered.

12Four numbers to understand before participating

Display	Meaning	Common misreading
Runtime	Cumulative compute time of validated tasks	Neither wall-clock study duration nor quantity of scientific outcomes
Results returned	Returned result instances counted by the statistics; replication can create several per workunit	Not unique workunits or experiments, and compute costs differ
WCG points / BOINC credit	Contribution accounting shown as 7 WCG points per BOINC credit	Not a currency issued by WCG, ownership, or correctness of the science
Devices / members	Cumulative registered devices and accounts	Not the number currently online

Participation runs from account registration and project selection to installing BOINC — or attaching an existing client — and setting a device profile. Use only devices whose owner has given permission, and begin with conservative limits on CPU share, hours, battery, network, and temperature. Spare compute still consumes electricity and may add heat and fan noise; hardware effects depend on settings, cooling, and device. WCG handles IP addresses, device specifications and performance, disk information, and preferences alongside account data; some member, device, and contribution statistics may be public depending on privacy settings.
There is no universally correct project or contribution level; compare research purpose, publication policy, and device cost yourself. Runtime, points, and result totals cannot by themselves yield electricity use or carbon emissions. Marginal device power, local grid mix, cooling, time powered on only for BOINC, replication, and retries would all have to be measured before claiming that idle-resource use has zero environmental cost.
WCG workunit dispatch can look like a mining pool sending hashing jobs to workers. But the Bitcoin protocol itself — and solo mining — does not receive jobs from a central server. WCG executes centrally selected scientific applications, validates application-specific outputs, and grants credit. In Bitcoin, miners propose candidate blocks and each full node independently applies the same consensus rules. WCG history teaches not that dispatching computation automatically decentralizes a system, but that purpose, authority, validation, and reward must be designed separately.

World Community Grid — Citizen Computing for Science

01Not one experiment, but shared infrastructure for public-interest research

02Grid, BOINC, and WCG are not synonyms

03From the Smallpox Grid to IBM infrastructure at world scale

04Choose the research before gathering compute

05How one workunit becomes research data

06Replication, trusted devices, and signatures divide trust

07Three active projects in August 2026

08From MCM to MAM — distributing combinatorial search

09Africa Rainfall Project — storage can bottleneck before compute

10“Completed” does not mean a finished medicine

11From IBM to Krembil — continuity required an operational transfer

12Four numbers to understand before participating

Primary sources

Read next

Citation / 引用情報

World Community Grid — Citizen Computing for Science

01Not one experiment, but shared infrastructure for public-interest research

02Grid, BOINC, and WCG are not synonyms

03From the Smallpox Grid to IBM infrastructure at world scale

04Choose the research before gathering compute

05How one workunit becomes research data

06Replication, trusted devices, and signatures divide trust

07Three active projects in August 2026

08From MCM to MAM — distributing combinatorial search

09Africa Rainfall Project — storage can bottleneck before compute

10“Completed” does not mean a finished medicine

11From IBM to Krembil — continuity required an operational transfer

12Four numbers to understand before participating

Primary sources

Read next

Read first

Related Topics

Go deeper

Citation / 引用情報