Mechanistic links
Mechanistic links turn a stack of papers into a structured map of cause and effect. As the agent reads across your selected literature, it extracts the molecular and biological relationships buried in methods and results sections and lays them out as a single, citation-grounded table you can scan, sort, and reuse.
#Mapping mechanisms across papers
A single paper usually describes a handful of relationships in prose — a factor that upregulates a target, a pathway that drives a phenotype, an inhibitor that blunts an outcome. Those relationships are the load-bearing claims of the literature, but they are scattered across figures, results paragraphs, and supplementary text, and no two papers state them the same way. Reading ten papers to reconstruct one mechanism by hand is slow and error-prone.
When you ask for mechanistic links, the agent reads the full text of each selected paper and pulls out the discrete cause-and-effect relationships it describes. It normalises them into a common shape — a source factor, the kind of link, the target it acts on, and the direction of the effect — so that a claim written one way in one paper and another way in a second paper line up in the same table. The result is a cross-paper synthesis that surfaces connections no single study makes explicit: an upstream trigger reported in one paper feeding a downstream outcome measured in another.
Because the extraction works from full text rather than abstracts, it captures the mechanistic detail that abstract-only tools miss. Abstracts state conclusions; methods and results sections state the interactions, magnitudes, and conditions that make a mechanism real.
#The mechanistic links table
Each row in the table is one mechanistic relationship. The columns are designed to be read left to right as a sentence — this factor acts on this target, in this direction, in this context — and to be grounded by the citation that supports it.
| Column | What it holds |
|---|---|
| Factor | The source of the relationship — a gene, protein, metabolite, drug, or other upstream entity that drives the effect. |
| Link type | The kind of interaction, such as activation, inhibition, gene regulation, binding, or a signalling event. |
| Target | The downstream entity the factor acts on — another molecule, a pathway, a cell type, or a measured phenotype. |
| Effect direction | Whether the factor increases, decreases, or otherwise modulates the target (for example ↑, ↓, or a described modulation). |
| Biological context | The conditions under which the link holds — the tissue, model system, disease setting, or experimental condition reported by the source. |
| Verified PMID | The PubMed identifier(s) of the paper(s) that support the link, so every row traces back to a specific source. |
Reading a row as a whole gives you a complete, self-contained claim: factor → link type → target, with the direction of the effect, the setting it was observed in, and the citation that backs it. Rows that share a factor or a target let you trace chains upstream and downstream across the whole set of papers.
#The Directness classification
Not every link is stated with the same directness by its source, so each row carries an evidence classification that tells you how the relationship was established. This lets you weigh the table rather than take every row at face value.
- Direct — the link is stated explicitly and measured in the source paper. The factor, target, and effect direction all come from what that study reports directly.
- Derived — the link is inferred by connecting statements, often across more than one paper. A derived row bridges an upstream claim in one place to a downstream claim in another, or reconstructs a relationship from separately reported pieces.
Direct links are the strongest evidence in the table; derived links are where the cross-paper synthesis is most valuable but also where you should look most closely at the underlying sources before treating a connection as established. The classification sits next to each row so you can filter your reading by confidence.
#Citation grounding
Every mechanistic link is tied to the specific source that supports it — there are no free-floating claims in the table.
- Each row carries the source PMID(s) so the connection can be traced back to the exact paper that supports it.
- The evidence type (Direct or Derived) is attached to each row so you can gauge the strength of the connection at a glance.
- Because links are drawn from full-text analysis, they capture mechanistic detail — the specific interaction, the direction, the conditions — that abstract-only extraction would miss entirely.
Grounding every row in a PMID means the table is not just a summary you have to trust; it is a set of pointers you can check. You can jump from any relationship straight to the passage that produced it.
#Use cases and plan allowances
A structured, citation-backed mechanism table is useful anywhere you would otherwise reconstruct relationships by hand:
- Pathway mapping — assemble the molecular mechanisms underlying a biological process from your literature set into one coherent view.
- Drug-target identification — surface potential intervention points by reading the upstream and downstream relationships around a target of interest.
- Grant proposals — include structured mechanistic summaries backed by peer-reviewed citations rather than hand-curated prose.
- Literature reviews — synthesise mechanistic knowledge from dozens of papers into a single sortable table instead of a narrative you have to write from scratch.
#Plan allowances
Mechanistic links are available on every plan. Each plan includes a monthly allowance that scales with tier — entry plans include a smaller allowance and higher plans include progressively larger ones. Paid plans also support pay-as-you-go generation if you exceed your monthly allocation, so heavier synthesis work is never hard-blocked. See the pricing page for the current per-tier figures.