Educational research platform

Before you begin

Welcome to Peptide Relay

Explore source-linked research, practical tools, and Community Intelligence with evidence, interpretation, and personal experience kept clearly separated.

Please confirm
Choose how you enter RelayYou can change your mind and create a workspace later.

No account is required for the Research Library, Learn, Compare, Stack Explorer, Community, or Tools. A free workspace is only required for private features such as My Relay, Protocol Tracker, saved work, following compounds, and managing Community Experiences.

Peptide Relay status

Public Alpha

v0.9.0

Building Relay with the community.

Relay is now feature-complete and has entered its first Public Alpha.

From this point forward, improvements are driven by real-world usage, community feedback, analytics, and research rather than internal feature planning.

Thank you for helping shape Relay.

What's NewWhat shipped in the current release
v0.9.0 — Public AlphaJuly 2026

Research Library

Thirty-five source-traceable compound and blend profiles with plain-English summaries, detailed science, evidence context, timelines, and references.

Learn

Peptide 101 and seven cornerstone guides for reading studies, mechanisms, evidence strength, personal experiences, and research uncertainty.

Compare

Side-by-side research context with shared, different, and unknown states—without inventing head-to-head conclusions.

Stack Explorer

Multi-compound pathway and evidence analysis with exact-combination limits and unanswered questions kept visible.

Community Experiences

Anonymous structured Experience Reports, separate story consent, verified follow-ups, moderation, and privacy-thresholded Community Intelligence.

Protocol Tracker

Private schedules, administrations, reflections, measurements, inventory, imports, lifecycle controls, and reconstitution support.

Reconstitution Hub

Single-compound and blend calculations, visual syringe and vial interpretation, reference marks, and private saved workspaces.

Relay-wide Search

One alias-aware search experience across public research and signed-in workspace destinations.

Mobile Optimization

Reliable touch selection, tighter information density, responsive tables and tabs, and mobile-first workflow refinement.

Motion System

One restrained motion language that explains state changes and respects reduced-motion preferences.

Platform Improvements

Database contract auditing, private-route indexing boundaries, public-route smoke tests, clearer recovery messages, and stronger protection against false-success writes.

RoadmapDirection without promised timelines

Roadmap items describe current direction. Public Alpha evidence may change their order or scope.

Now
  • Community growth
  • Bug fixes
  • UX improvements
  • Content expansion
Next
  • Relay+ features
  • Additional research tools
  • Expanded compound library
Future
  • Relay AI
  • Native iPhone app (planned)
  • Native Android app (planned)
Known IssuesMeaningful issues users may encounter
No known critical issues at this time.

Newly confirmed issues will appear here when they meaningfully affect the public experience.

FeedbackHelp improve the next release

Found a bug?Have a suggestion?

Submit feedback to help improve Relay
Version HistoryPublic releases and milestones
v0.9.0

Public Alpha

The first feature-complete public release candidate for Peptide Relay.

Released July 2026

Last updated July 2026 · Updated with every public release.

Learn · Judging evidence

How strong is this evidence?

Learn why evidence strength depends on the exact claim, design fit, study integrity, uncertainty, and the complete body of evidence.

The simple answer

Evidence is strong when it gives us good reason to trust a specific conclusion—not merely when it carries an impressive study label. The design must fit the question, the study must protect against important bias, the measurements must match the claim, and the result must be precise enough to rule out meaningfully different explanations.

Confidence usually comes from the whole body of relevant evidence. Agreement across well-conducted, sufficiently direct studies strengthens a conclusion; serious bias, indirectness, imprecision, inconsistency, and missing results weaken it. The same evidence can be strong for one claim and weak for a broader one.

7 min read Reviewed July 28, 2026 8 sources

Begin with the exact claim

Evidence strength is not a permanent score attached to a paper. It is a judgment about how confidently a particular result or body of evidence supports a particular claim.

A study may provide strong evidence that a compound binds a target in a laboratory system and weak evidence that it improves a person-relevant outcome. Nothing about the study changed; the claim did. Before asking how strong the evidence is, define the population or system, exposure, comparison, outcome, and timeframe the claim actually covers.

Study design sets possibilities—not a final grade

Different designs protect against different errors and answer different questions. Randomization can make comparison groups more alike at the start and strengthen causal inference about an assigned intervention. Observational studies may reveal long-term patterns, uncommon harms, or real-world associations that a trial was not designed or large enough to capture. Laboratory and animal studies can investigate mechanism under controlled conditions.

Those strengths are question-specific. A randomized trial is not automatically informative about an outcome it never measured. An observational study does not become useless because it cannot remove every confounder. A design label tells us which conclusions may be possible and which risks deserve attention; it does not complete the appraisal.

Execution determines whether the design delivered

A good design can be weakened by poor allocation concealment, large or unequal loss to follow-up, deviations from the planned intervention, unreliable outcome measurement, or selective choice among several analyses. Cochrane therefore assesses risk of bias for a specific result—not simply for the publication as a whole.

The same trial can have different risk-of-bias judgments for different outcomes. An objectively recorded laboratory result may be less vulnerable to awareness of treatment assignment than a subjective outcome in an unblinded study. Appraisal has to follow the result being used.

Directness asks how far the evidence must travel

Direct evidence closely matches the question of interest. Differences in species, population, health state, exposure, comparator, measurement, route, dose range, or follow-up can make evidence indirect for a particular conclusion.

Indirect does not mean irrelevant. It means an extra inference is required. A short trial may directly answer a short-term biomarker question while remaining indirect for durable functional benefit or rare harm. The larger the distance between what was studied and what is claimed, the more uncertainty the conclusion must carry.

Precision asks which effects remain compatible with the data

A point estimate is only one value supported by a dataset. A confidence interval describes a range of effects compatible with the result under the statistical model. A wide interval may include meaningful benefit, little difference, and meaningful harm at the same time.

A small p-value does not measure effect size, practical importance, or the probability that a hypothesis is true. Strength depends on the estimate, its uncertainty, the study design and assumptions, and whether the remaining range would lead to materially different interpretations.

Consistency matters—but counting studies is not enough

When independent studies using suitable methods produce compatible findings, confidence may increase. When results differ, the disagreement should be investigated rather than averaged away. Population, exposure, duration, outcome definition, bias, or chance may explain part of the variation.

Five small studies with the same flaw do not necessarily outweigh one stronger study. Repetition strengthens evidence when it provides genuinely informative, reasonably independent tests—not merely more copies of the same limitation.

The body of evidence is stronger than a winner-take-all paper

Systematic review methods aim to identify all relevant studies, assess their limitations, and synthesize results transparently. Frameworks such as GRADE then judge certainty for each important outcome using domains including risk of bias, inconsistency, indirectness, imprecision, and publication bias.

A systematic review is not automatically high-certainty evidence. Its conclusion depends on the search, eligibility decisions, included studies, analytic methods, and whether missing results could distort the available record. Synthesis can organize uncertainty; it cannot manufacture evidence that the underlying studies do not contain.

Key terms

Evidence strength
How much confidence the available evidence supports for a specific conclusion in a defined context.
Risk of bias
The possibility that features of a study's design, conduct, analysis, or reporting systematically distorted a result.
Directness
How closely the studied population or system, exposure, comparison, outcome, and timeframe match the question of interest.
Precision
How narrowly the evidence estimates an effect and how many meaningfully different effects remain compatible with the data.
Inconsistency
Important unexplained variation in the direction or size of results across studies.
Publication bias
Distortion that occurs when the availability of results depends partly on what the study found.
Certainty of evidence
A structured judgment about confidence in a result or range of effects for a specific outcome.

Evidence strength is built claim by claim

Exact claimWhat population, exposure, comparison, outcome, and timeframe?
Design fitCan this design answer that kind of question?
Study integrityCould bias have systematically changed this result?
Result uncertaintyWhich meaningfully different effects remain compatible?
Body of evidenceIs the full record direct, consistent, and complete enough?
Bounded conclusionHow far can confidence reasonably travel?
Each stage asks a different question. An impressive design cannot repair a mismatched claim, serious bias, unresolved uncertainty, or a distorted evidence base.

What this can—and cannot—tell us

What it can tell us

  • How confidently the available evidence supports a clearly defined claim.
  • Which parts of a conclusion are directly measured and which require extrapolation.
  • Whether important sources of bias could plausibly change a specific result.
  • How much uncertainty remains around the direction and size of an effect.
  • Whether results are reasonably consistent across relevant, independent studies.
  • Why confidence should increase, decrease, or remain limited for each outcome.

What it cannot establish alone

  • That every conclusion in a highly ranked study design is strong.
  • That a statistically significant result is large, important, unbiased, or reproducible.
  • That several studies outweigh one study merely because there are more of them.
  • That a systematic review is reliable without appraising its methods and included evidence.
  • That evidence strong for one outcome, population, or timeframe supports a broader claim.
  • What decision a person should make; evidence certainty and recommendations are separate judgments.
Go deeperOptional · about 2 minutes

Why p < 0.05 is not an evidence-strength meter

A p-value describes how incompatible the observed data are with a specified statistical model, assuming the model and its null hypothesis. It does not report the probability that the hypothesis is true, the size or importance of an effect, or whether the study was free from bias.

Scientific interpretation therefore needs effect estimates, uncertainty intervals, design and conduct, outcome relevance, transparency, and the wider evidence. Moving a p-value from one side of a threshold to the other does not transform weak reasoning into strong evidence.

Why a systematic review can still leave very low certainty

A review may use excellent methods yet find only small, indirect, biased, inconsistent, or selectively reported studies. In that case, confidence in the review process can be high while certainty in the effect estimate remains low.

The review tells us the evidence base is limited more reliably; it does not make the underlying effect more certain. That distinction is one reason GRADE evaluates certainty by outcome rather than assigning one quality label to the entire article.

Certainty is not the same thing as a recommendation

Evidence certainty addresses confidence in what is likely to happen. A recommendation also depends on the magnitude of benefits and harms, values and preferences, feasibility, resources, equity, and the decision context.

High-certainty evidence can support a small or undesirable effect. Low-certainty evidence may still describe a serious possibility worth investigating. Relay keeps the evidence judgment separate from telling a person what to do.

The takeaway

If you only remember one thing from this guide:

Evidence is only as strong as its fit to the claim, its protection against bias, and the uncertainty that remains.

Sources and support8 sources
  1. Defines certainty in relation to a range or decision threshold and identifies the domains used to appraise a body of evidence.

  2. Explains outcome-specific GRADE assessment across risk of bias, inconsistency, indirectness, imprecision, and publication bias.

  3. Shows why risk of bias is assessed for a specific result across domains of trial design, conduct, and reporting.

  4. Describes confounding and other biases in non-randomized intervention studies and the result-specific ROBINS-I approach.

  5. Explains confidence intervals, study weighting, inconsistency, missing data, and the limits of statistical synthesis.

  6. Addresses certainty, applicability, effect interpretation, and the need to keep conclusions within the evidence.

  7. Explains why p-values do not measure effect size, importance, or the probability that a hypothesis is true and should not alone determine conclusions.

  8. Provides the current minimum reporting standards needed to appraise how a randomized trial was designed, analyzed, and interpreted.