The key to rapid and accurate literature curation for clinical interpretation and therapeutic discovery is standardizing and optimizing the review process. The Mastermind Genomic Search Engine meets this need with a collection of sophisticated algorithms that prioritize search results by relevance. This proprietary process is called the Mastermind Relevancy Score.
Why is the Mastermind Relevancy Score Useful?
The best way to speed up literature curation for genes and variants without sacrificing comprehensiveness is to reduce the number of articles necessary to achieve a sufficient literature review.
By reading the most relevant articles first in Mastermind, each subsequent decision can be focused on filling in the blanks, and decisions can be reached sooner. The quicker information is found, the fewer articles need to be read, which saves time and increases productivity and throughput.
How is the Mastermind Relevancy Score Determined?
Mastermind Relevancy Score is a set of algorithms used to prioritize (or “score”) articles by their relevance to a search and present the most relevant articles to the user. Mastermind calculates this score in two constituent parts:
1. Clinical Relevancy Score
This first calculation is the more objective measure, irrespective of context. It seeks to assign a baseline likelihood that the article is relevant to anyone looking for genomic information, regardless of the particular genes, variants, diseases, or other filters in their search (for a list of available filters, or associations, see Mastermind’s Genomic Associations). This incorporates numerous parameters intrinsic to each article itself, including:
- The type of journal in which the article was published
- The Journal Impact Score (or Citation Factor) for this publication
- The inclusion of information specific to genomics studies, such as cohort studies, etc.
- Recency, or date of publication
2. Contextual Relevance
The second part of the relevancy calculation is the contextual relevance to the current search being performed. This includes calculations which answer the following questions.
How likely is the article to be talking about the searched gene or variant, as opposed to mentioning them incidentally?
For example, if a variant appears more frequently throughout the article compared to other articles, then the article will be more highly prioritized. Likewise, if a variant appears more frequently within an article than other variants mentioned in that same article, it’s more likely to be talking about the frequently mentioned variant and incidentally mentioning the other variants.
Another method for determining whether an article is written with a focus on a given gene, variant, or other association is to consider where in the article the searched associations are mentioned. An article that mentions a gene in the title is more relevant than one mentioning the gene elsewhere in the abstract, introduction, or methods, etc., which, in turn, is more relevant than an article that only mentions the gene in the full text, which, in turn, is more relevant than an article that only includes the gene in the supplemental data.
Mastermind is also able to classify and distinguish between tables, figures, and sections such as the bibliography to weigh the significance of or exclude terms entirely as appropriate to optimize specificity.
How strongly does the article associate the searched genes, variants, and other associations?
For example, if a disease and variant are mentioned together in the title, the two words are more strongly associated than if the disease is in the title and the variant is mentioned further down in the text. In the case of a multi-parameter search such as looking for two individual variants, the closer these terms are to each other within the article increases the prioritization for the article in the final results list.
This is a useful measure of the relevance of the result, since scientific articles may describe multiple things in different sections of the text. Placing a greater emphasis on the prioritization of articles that mention the searched terms in the same sentence, the same paragraph, or the same section of the article as opposed to mentioning the terms in widely separated sections of the article improves the specificity and relevance of the more highly prioritized results.
The final search results are then prioritized by combining the objective and contextual relevancy scores to list the most clinically relevant articles for your search.
How Can Mastermind Prioritization be Tailored to My Search?
Article prioritization can be tailored to your use-case by adding terms to your search to enhance the specificity of the results. The simplest way to do this is to add a specific disease or phenotypes to your search. The searched-for disease or phenotypes then become terms used to filter and enhance the prioritization of results.
Another way to enhance the specificity of your search results is to add terms from the categories provided just below the search bar, as indicated in the screenshot below. This catalog of terms is organized by categories, which themselves include further sub-categories.
Sub-categories include terms for ACMG/AMP variant interpretation (such as PS3/BS3 Functional Significance or PM4 Segregation studies) as well as terms to help determine the meaningfulness of somatic variants based on Diagnostic/Prognostic or Therapeutic relevance in the Clinical Significance category.
Finally, identifying articles that discuss structural alterations such as fusion genes is also possible using the Genetic Mechanism category, while other structural alterations like CNVs can be searched directly.
How Else Can Articles Be Sorted in Mastermind?
In addition to the default Relevance calculation which our users find most useful, the articles list can be sorted in the following ways.
- Article Publication Date – This shows the most recently published articles first, making it easy to stay up to date with the latest information published for your search. If this is frequently useful for your workflow, consider signing up for Mastermind Alerts to receive notifications with new results as they’re published on a weekly, monthly, or quarterly basis.
- Journal Impact Score – Also known as Citation Factor, this feature shows the articles in the most cited journals first, which can be a useful proxy for finding articles that tend to be more impactful.
- Number Gene or Variant Matches – This shows the articles which mention the searched gene or variant the most, which is sometimes useful as a highly simplified prioritization metric.
These sort functions can all be selected from the drop-down as highlighted in the image below, at the top right of the Articles list pane.
Based on the experience of our users, the first results in Mastermind are sufficient in >90% of cases to make a complete and accurate variant interpretation, resulting in a quicker and more comprehensive variant curation process.
The Mastermind Relevancy Score is a small part of why variant scientists can count on Mastermind as the first resource for variant literature interpretation. We welcome your thoughts at firstname.lastname@example.org.
Begin prioritizing search results by relevance, and experience this feature for yourself. Sign up for Mastermind here.
Are you using Mastermind Basic Edition? See the most evidence and advanced features by upgrading to Professional Edition.