Assembling the evidence…
Biomarkers, as measurements of defined biological characteristics, can play a pivotal role in estimations of disease risk, early detection, differential diagnosis, assessment of disease progression and outcomes prediction. Studies of cancer biomarkers are published daily; some are well characterized, while others are of growing interest. Managing this flow of information is challenging for scientists and clinicians. We sought to develop a novel text-mining method employing biomarker co-occurrence processing applied to a deeply indexed full-text database to generate time-interval-delimited biomarker co-occurrence networks. Biomarkers across 6 cancer sites and a cancer-agnostic network were successfully characterized in terms of their emergence in the published literature and the context in which they are described. Our approach, which enables us to find publications based on biomarker relationships, identified biomarker relationships not known to existing interaction networks. This search method finds relevant literature that could be missed with keyword searches, even if full text is available. It enables users to extract relevant biological information and may provide new biological insights that could not be achieved by individual review of papers.
identified biomarker relationships not known to existing interaction networks
48/104 selected pairs were absent from named databases, but absence from databases is not full novelty proof
develop a novel text-mining method employing biomarker co-occurrence processing applied to a deeply indexed full-text database
workflow is implemented with specified corpus, thresholds, networks, clustering, and validation examples
Biomarkers across 6 cancer sites and a cancer-agnostic network were successfully characterized
networks and selected biological contexts are reported across sites, though detailed validation covers selected clusters
This search method finds relevant literature that could be missed with keyword searches
claim is plausible from relationship search design but lacks direct keyword-search comparator testing
this approach is not limited to oncology
generalizable in principle, but only oncology seed lists and cancer networks are evaluated
Derived from the full evaluation — not a separate score.
Strengths
The article specifies the EDRN biomarker seed set, Dimensions full-text corpus, date range, proximity rules, co-occurrence thresholds, edge weighting, clustering method, and enrichment analysis. This makes the workflow assessable even though full reproducibility is limited by missing code and proprietary corpus access.
↳ Methods, p. 2–4
The method is applied to six cancer-site networks and a cancer-agnostic network, rather than a single illustrative example. The results then examine renal, colorectal, and cancer-agnostic clusters with different contextual-validation approaches.
↳ Abstract; Results, p. 3–8; Figures 2–9
The Discussion identifies the difficulty of optimizing co-occurrence parameters and acknowledges that the method does not identify relation type, molecular nature, or negation. These limitations are concrete and tied to plausible future improvements such as sentence-level processing and context-aware models.
↳ Discussion, p. 14–15
Limitations
The Abstract and Conclusion state that the method finds relevant literature missed by keyword searches, but the paper does not report a head-to-head comparison against keyword-search recall or precision. This limits the strength of the claimed retrieval advantage.
↳ Abstract; Conclusion, p. 14–15
Manual validation focuses on three selected example analyses, including fastest-growing clusters and top-Mendeley publications. These examples support plausibility but do not establish systematic performance across the full 31,550-pair co-occurrence network.
↳ Methods, Contextual analysis, p. 3; Results, p. 4–14
The paper names packages and states that networks are published on NDEx, but it does not provide executable analysis code or full access to the underlying Dimensions corpus. This matters for a methods paper whose results depend on preprocessing, thresholding, and proprietary full-text indexing.
↳ Methods, p. 2–3
The article’s strongest scores come from a clear, assessable workflow and a coherent presentation of how full-text co-occurrence networks can surface biologically plausible biomarker relationships. The main methodological constraint is validation scope: the paper manually reviews selected high-growth clusters and Mendeley-ranked papers, but does not evaluate performance across the full network or against keyword-search baselines. The contribution is therefore best read as a useful proof-of-concept literature-mining tool rather than a validated retrieval system. Impact potential is moderate because the open-access tool and NDEx-hosted networks create a plausible route to use, while operational uptake and decision impact remain untested.
Moderate3.4/5.0
The paper contributes a practical full-text co-occurrence and network-analysis workflow for oncology biomarkers, applied across six cancer sites and a cancer-agnostic network. Its contribution is meaningful but incremental because co-occurrence text mining is established and the comparative advantage over keyword search is not directly tested.
identified biomarker relationships not known to existing interaction networks
The methods report the corpus, seed list, proximity window, co-occurrence thresholds, network construction, Leiden clustering, publication-growth metric, and enrichment testing. Methodological Rigour is limited by ad hoc parameter choice, selected-cluster manual validation, proprietary Dimensions dependence, and no public analysis code.
ad hoc analysis indicated that a 10-word window is slightly too short
The article follows a coherent sequence from motivation to search strategy, network generation, selected validation examples, enrichment analysis, and limitations. Figures and tables support comprehension, although some abstract and conclusion language states the retrieval advantage more strongly than the reported benchmark evidence allows.
The results of our searches and network analysis are summarized
The paper positions itself against prior abstract-only, mutation-focused, and open-access-limited text-mining approaches, and it contextualizes selected biomarker clusters with cancer-biology literature and pathway databases. Contextual depth is moderated by limited engagement with formal information-retrieval benchmarking and relation-extraction alternatives.
these approaches, which are limited in scope
Watch-outs4 of 4 checks
The paper’s methods-to-results chain is broadly coherent, with consistent reported counts and a clear mapping from network construction to validation examples. Some headline claims extend beyond the evaluation design, especially claims about keyword-search superiority and minimal noise from selected validation subsets.
The study uses published literature and does not raise an apparent ethics-review requirement. Funding, sponsorship, employee affiliations, and consulting relationships are disclosed, with no visible conduct-integrity concern in the supplied text.
Flags: 2 declared / 5 total
39 references verified via PubMed, Crossref, or OpenAlex.
No retraction notice found in Retraction Watch.
Sources: Retraction Watch ✓PubPeer (coming soon)
Medium3.2/5.0
The paper addresses a recognized information-management problem for scientists and clinicians handling daily cancer-biomarker publications. Stakeholders and need are visible, but downstream decision settings such as guideline development or pipeline prioritization are not operationally evaluated.
managing this flow of information is challenging for scientists and clinicians
The workflow is implemented and linked to an interactive open-access tool, which gives it more translational readiness than a purely conceptual method. Readiness remains preliminary because the paper reports illustrative validation rather than user testing, workflow integration, or demonstrated decision change.
It is freely accessible at https://reports.dimensions.ai/mined-oncology-biomarkers/
Application across six cancer sites plus a cancer-agnostic network supports transferability within oncology. Broader transfer is constrained by the cancer-specific EDRN seed list, Dimensions corpus dependence, and validation concentrated in three selected clusters.
this approach is not limited to oncology
The work sits on a plausible cumulative path because it publishes networks, identifies biomarker pairs absent from selected interaction databases, and proposes temporal validation against later clinical emergence. The trajectory remains moderate because practical uptake and retrospective predictive validation are future work.
Further work could look to identify those biomarker pairs
Nabu’s assessment, alongside the field’s view.
Evaluation Metadata
AI-powered, human-governed. Something look off? Contact us to request a review.