5.18, 5.20 - Evaluating evidence and scientific conclusions

Genomes, proteomes and common ancestry

Inherited molecules provide evidence about ancestry because descendants inherit sequences from ancestors, while mutations can introduce changes over generations.

Genomics studies genomes, including comparisons of DNA sequences across organisms. Scientists compare corresponding regions and align them so that positions inherited from the same ancestral region can be compared. Shared sequence patterns, including shared changes, can support common ancestry. When suitable sequences and evolutionary rates are comparable, fewer differences generally support a more recent common ancestor. This does not mean that one living species evolved from the other.

Proteomics studies the proteins produced by cells or organisms, including their identities, sequences and abundance. For evolutionary relationships, scientists can compare the amino acid sequences of corresponding proteins across species. DNA coding sequences determine amino acid sequences, so inherited DNA changes can lead to protein differences. Some DNA changes do not change an amino acid because the genetic code is degenerate. Protein and DNA comparisons therefore need not show identical numbers of differences.

Consider these original illustrative data, not results from a real study. Each comparison uses the same aligned DNA region and the same corresponding protein region in three invented species. Only reliably identified positions are included.

Both comparisons support A and B as the closest pair in these markers. Their smaller differences fit a more recent common ancestor, with less divergence since their lineages separated. The agreement is useful, but a protein and its own coding gene are related evidence, rather than fully independent tests. Agreement across many suitable genes, proteins, fossils and other evidence is stronger than one short sequence alone.

Real studies illustrate how methods extend the evidence. Yoo and colleagues' 2025 ape-genome study compared six ape species and resolved previously difficult repetitive regions. Welker and colleagues' 2019 study recovered enamel protein sequences from an ancient Gigantopithecus tooth and supported a relationship closest to the orangutan lineage among living apes. Proteins can sometimes preserve evidence where usable ancient DNA is unavailable; that does not mean all proteins survive in all fossils.

New molecular findings can refine particular relationships or mechanisms within evolutionary theory. Support for evolution comes from the combined explanatory success of many tested lines of evidence, rather than the authority of one publication.

Testing molecular inferences

A molecular comparison is only as useful as the choice of sequences, the quality of measurement and the reasoning linking the result to ancestry.

Homologous features have a shared evolutionary origin. A useful sequence comparison identifies corresponding genes or proteins inherited from an ancestral sequence, rather than comparing unrelated molecules that happen to have similar functions. Gene duplication can leave several related copies in one genome; comparing different copies across species can give a misleading account of the species' relationships. Alignment also matters: missing positions and insertions must not make unlike positions appear equivalent.

Similarity alone is not conclusive. Convergent evolution can produce similar features independently under similar selection pressures. Selection can also favour the same amino acid changes independently in different lineages. Conversely, a protein essential to the same function may change very slowly in many lineages, making it poor at distinguishing close relatives. Several changes at one position can hide earlier changes. Researchers therefore consider the marker, the model of sequence change, the organisms sampled and agreement across different regions. A sequence difference count alone does not supply an exact date of divergence.

The following checks address different weaknesses; they are not interchangeable.

A blank showing no contamination is reassuring but cannot prove that every specimen is uncontaminated. Repeating the same analysis of the same contaminated sample can reproduce the same error. Independent material, controls and analysis are therefore valuable. When two methods disagree, scientists investigate whether their data, assumptions or evolutionary histories differ; they do not simply choose the result they prefer.

Community scrutiny and validation

Scientists make research inspectable so that a claim can be tested beyond the group that first proposed it.

A scientific journal provides a lasting, citable account of a study: its question, methods, results, analysis and conclusions. Detailed methods and accessible data let other scientists examine what was actually done. A research article reports an investigation; a review article brings together and evaluates existing investigations. Reading the methods and original evidence is more informative than relying on a striking headline or a count of citations.

In peer review, other researchers with relevant expertise critically assess a submitted paper. Reviewers can identify inappropriate comparisons, missing controls, weak analysis or conclusions that exceed the results. Authors may have to explain choices, supply further evidence or revise the paper; editors may reject it. Reviewers should declare relevant competing interests, and editors should manage them. The Royal Society describes its editorial independence and measures to reduce review bias.

Peer review is a quality check, with limits. Reviewers do not normally repeat the entire study and may miss errors, undisclosed problems or fraud. Publication therefore makes a claim available for wider scrutiny; it does not guarantee truth. A prestigious journal or an impressive job title cannot replace inspection of the evidence.

At scientific conferences, researchers present findings, answer questions and discuss methods with specialists. A question about contamination might lead to a better control; disagreement about alignment might lead to a new comparison. Meetings also allow collaboration and rapid exchange of developing ideas. A conference presentation may be preliminary and has not necessarily undergone the same review as a full journal article. Applause, attendance or publicity is not validation.

Scrutiny continues after publication. Other groups can reanalyse shared data and code, collect new samples, repeat a method or test a prediction using a different method. Obtaining consistent results with the original data and analysis checks computational reproducibility; collecting independent data to test the same question checks whether the finding is replicated. Terminology varies across fields, so state what was repeated. The National Academies makes this distinction explicitly. Agreement within expected uncertainty increases confidence. Failure to agree prompts investigation and may lead to correction, retraction or a revised explanation; one failure needs interpretation rather than automatic dismissal of all related evidence.

For evolutionary research, sharing the aligned sequences, raw measurements, sample histories and analysis choices lets the community test a proposed relationship. The fossil-enamel study above makes protein data available through a public repository. This enables checking; it does not establish that every possible independent check has already succeeded. Evolution is accepted because converging evidence withstands such tests over time, rather than because scientists vote a theory into truth.