Ten quick tips for biomarker discovery and validation analyses using machine learning

Dı́az-Uriarte, Ramón; Lope, Elisa Gómez de; Giugno, Rosalba; Fröhlich, Holger; Nazarov, Petr V.; Nepomuceno-Chamorro, Isabel A.; Rauschenberger, Armin; Glaab, Enrico

doi:10.1371/journal.pcbi.1010357

Cited by 14 publications

(14 citation statements)

References 141 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…through projects like EU-STANDS4PM [ 28 ], which are in the process of developing an ISO document (ISO/AWI TS 9491-1) on translational standards for these models. The PERMIT project has also addressed this issue through a scoping review and recommendations [ 6 , 29 ]. The success of such efforts is also dependent upon the development of a global translational medicine community to coordinate interdisciplinary research that can better address unmet medical needs.…”

Section: Discussionmentioning

confidence: 99%

Recommendations for robust and reproducible preclinical research in personalised medicine

Fosse

Oldoni²,

Bietrix³

et al. 2023

BMC Med

Self Cite

View full text Add to dashboard Cite

Background Personalised medicine is a medical model that aims to provide tailor-made prevention and treatment strategies for defined groups of individuals. The concept brings new challenges to the translational step, both in clinical relevance and validity of models. We have developed a set of recommendations aimed at improving the robustness of preclinical methods in translational research for personalised medicine. Methods These recommendations have been developed following four main steps: (1) a scoping review of the literature with a gap analysis, (2) working sessions with a wide range of experts in the field, (3) a consensus workshop, and (4) preparation of the final set of recommendations. Results Despite the progress in developing innovative and complex preclinical model systems, to date there are fundamental deficits in translational methods that prevent the further development of personalised medicine. The literature review highlighted five main gaps, relating to the relevance of experimental models, quality assessment practices, reporting, regulation, and a gap between preclinical and clinical research. We identified five points of focus for the recommendations, based on the consensus reached during the consultation meetings: (1) clinically relevant translational research, (2) robust model development, (3) transparency and education, (4) revised regulation, and (5) interaction with clinical research and patient engagement. Here, we present a set of 15 recommendations aimed at improving the robustness of preclinical methods in translational research for personalised medicine. Conclusions Appropriate preclinical models should be an integral contributor to interventional clinical trial success rates, and predictive translational models are a fundamental requirement to realise the dream of personalised medicine. The implementation of these guidelines is ambitious, and it is only through the active involvement of all relevant stakeholders in this field that we will be able to make an impact and effectuate a change which will facilitate improved translation of personalised medicine in the future.

show abstract

Section: Discussionmentioning

confidence: 99%

Recommendations for robust and reproducible preclinical research in personalised medicine

Fosse

Oldoni²,

Bietrix³

et al. 2023

BMC Med

Self Cite

View full text Add to dashboard Cite

show abstract

“…It is important to keep in mind that while features derived from other machine learning prediction tools are usually more comprehensive, they might not be as accurate as observations, e.g., from mass spectrometry studies. However, to circumvent annotation bias and lost interpretability, , we actively decided against other strategies as the research aim included biological interpretation. Nevertheless, as a deep learning-based predictor did not perform better than our model, the prediction task investigated here might not warrant the use of such sophisticated models because of the limited amount of data available.…”

Section: Discussionmentioning

confidence: 99%

“…Thus, if gaining biological insights from a prediction model is an additional research aim, use of "shallow" but interpretable models might be more beneficiary. 15 Accordingly, we were interested in developing a CSF secretion predictor considering two important aspects: 1) By limiting model training to likely CNS-originating proteins, our study focuses on specifically distinguishing between proteins secreted from the brain to CSF and those confined to the brain; 2) We prioritize model interpretability by utilizing an explainable machine learning model and biologically informative features to investigate how the model makes its decisions and to explore the biology behind CSF secretion. Note that while we refer here and throughout this study to protein secretion, this term is meant to include all physiological processes that lead to a brain protein's presence in CSF.…”

Section: ■ Introductionmentioning

confidence: 99%

“…While reporting high model accuracy, biological insight from these models is very limited; the infamous “black box” character of deep learning hinders us to understand how the model makes its decisions. Thus, if gaining biological insights from a prediction model is an additional research aim, use of “shallow” but interpretable models might be more beneficiary …”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deciphering Protein Secretion from the Brain to Cerebrospinal Fluid for Biomarker Discovery

Waury,

de Wit,

Verberk

et al. 2023

J. Proteome Res.

View full text Add to dashboard Cite

Cerebrospinal fluid (CSF) is an essential matrix for the discovery of neurological disease biomarkers. However, the high dynamic range of protein concentrations in CSF hinders the detection of the least abundant protein biomarkers by untargeted mass spectrometry. It is thus beneficial to gain a deeper understanding of the secretion processes within the brain. Here, we aim to explore if and how the secretion of brain proteins to the CSF can be predicted. By combining a curated CSF proteome and the brain elevated proteome of the Human Protein Atlas, brain proteins were classified as CSF or non-CSF secreted. A machine learning model was trained on a range of sequence-based features to differentiate between CSF and non-CSF groups and effectively predict the brain origin of proteins. The classification model achieves an area under the curve of 0.89 if using high confidence CSF proteins. The most important prediction features include the subcellular localization, signal peptides, and transmembrane regions. The classifier generalized well to the larger brain detected proteome and is able to correctly predict novel CSF proteins identified by affinity proteomics. In addition to elucidating the underlying mechanisms of protein secretion, the trained classification model can support biomarker candidate selection.

show abstract

“…Biomarker development typically begins with quantification of target molecules in samples from a discovery cohort of people with disease and appropriate control individuals. Positive findings are subsequently followed up by detecting potential biomarkers in samples from independent replication and/or validation cohorts [5,[18][19][20]. Once validated, biomarker integration into existing care pathways requires identifying, optimizing and streamlining suitable detection methodologies.…”

Section: Biomarker Developmentmentioning

confidence: 99%

Identification and detection of microRNA kidney disease biomarkers in liquid biopsies

Smith,

Redman,

Fraser

et al. 2023

Current Opinion in Nephrology &Amp; Hypertension

View full text Add to dashboard Cite

Purpose of review MicroRNAs (miRNAs) are emerging rapidly as a novel class of biomarkers of major organ disorders, including kidney diseases. However, current PCR-based detection methods are not amenable to development for high-throughput, cost-effective miRNA biomarker quantification. Recent findings MiRNA biomarkers show significant promise for diagnosis and prognosis of kidney diseases, including diabetic kidney disease, acute kidney injury, IgA nephropathy and delayed graft function following kidney transplantation. A variety of novel methods to detect miRNAs in liquid biopsies including urine, plasma and serum are being developed. As miRNAs are functional transcripts that regulate the expression of many protein coding genes, differences in miRNA profiles in disease also offer clues to underlying disease mechanisms. Summary Recent findings highlight the potential of miRNAs as biomarkers to detect and predict progression of kidney diseases. Developing in parallel, novel methods for miRNA detection will facilitate the integration of these biomarkers into rapid routine clinical testing and existing care pathways. Validated kidney disease biomarkers also hold promise to identify novel therapeutic tools and targets. Video abstract http://links.lww.com/CONH/A43

show abstract

Ten quick tips for biomarker discovery and validation analyses using machine learning

Cited by 14 publications

References 141 publications

Recommendations for robust and reproducible preclinical research in personalised medicine

Recommendations for robust and reproducible preclinical research in personalised medicine

Deciphering Protein Secretion from the Brain to Cerebrospinal Fluid for Biomarker Discovery

Identification and detection of microRNA kidney disease biomarkers in liquid biopsies

Contact Info

Product

Resources

About