Xuzeng He

An introduction to tools supporting PubMed

<figure>
<img src="https://cdn-images-1.medium.com/max/894/0*wa8pouqfyGYwIZ2G" />
<figcaption>Source: Generated using DeviantArt’s DreamUp</figcaption>
</figure>

### Author:

- Xuzeng He (**ORCID:**
  [0009--0005--7317--7426](https://orcid.org/0009-0005-7317-7426))

PubMed is a free Web literature search service developed and maintained
by the National Center for Biotechnology Information (**NCBI**), and it
is also a part of NCBI's Entrez retrieval system that provides access to
a diverse set of 38 databases. PubMed currently includes citations and
abstracts from over 5000 life science journals for biomedical articles
dating back to 1948. Since its inception, PubMed has served as the
primary tool for electronically searching and retrieving biomedical
literature. Every day, users worldwide issue millions of queries,
relying on its access to stay updated on the latest developments and
make discoveries in their respective fields.

In this post, we introduce some efforts made by various researchers to
make PubMed more useful in different directions.

### PubMed Labs

As the biomedical literature grows at an exponential rate, the National
Center for Biotechnology Information (NCBI) at the National Library of
Medicine has recently developed PubMed Labs, an experimental platform
for users to test new features/tools and provide feedback, which enables
us to make more informed decisions about potential changes to improve
the search quality and overall usability of PubMed.

PubMed Labs is implemented as a standalone service, separate from the
production operation of PubMed, so that it is non-disruptive to the
routine information seeking process of current PubMed users. It now has
several unique features that can distinguish it from PubMed. For
example, given a free-text query as input, search results are sorted by
Best Match in order to provide users with the most pertinent information
(in PubMed, the default sort order is Most Recent), which can provide
users with better results. PubMed Labs also has a more modern user
interface, allowing the user to discover relevant content in an
easier way.

<figure>
<img
src="https://cdn-images-1.medium.com/max/1024/0*mdHsqAevPZCSb8b2" />
<figcaption>PubMed Labs search results page with highlighted features.
(a) Search box. (b) Sort order toggle. (c) Query auto completion. (d)
Search facets. (e) Highlighted search terms in title and snippet. (f)
Related searches. (<a
href="https://arxiv.org/search/cs?searchtype=author&amp;query=Fiorini,+N">Nicolas
Fiori</a>ni et al., 2018) Article Link: <a
href="https://arxiv.org/abs/1806.04004">https://arxiv.org/abs/1806.04004</a></figcaption>
</figure>

### PubMed Knowledge Graph

In the field of healthcare and medicine, experts usually communicate
using medical jargon, which is compiled and stored in computer
processable collections such as [SNOMED CT](https://www.snomed.org/),
[ICD-10](https://www.cdc.gov/nchs/icd/icd10.htm),
[PubChem](https://pubchem.ncbi.nlm.nih.gov/), and [gene
ontology](https://geneontology.org/). These medical language terms
(e.g., genes, drugs, proteins, species, and mutations) are the backbone
of quality healthcare.

Conventional text mining tools are ineffective in handling this data
because these tools struggle to accurately interpret and extract
information from the intricate and domain-specific language used in
medical literature. The complexity and specificity of medical terms, as
well as the diverse ways they can be expressed in literature, pose
significant challenges for traditional text mining approaches. As a
result, the tools may fail to provide precise and comprehensive results,
making literature searches in healthcare and medicine more cumbersome.
Therefore, many studies have been devoted to building open-access
datasets to solve these bio-entity recognition problems.

One way to build an open-access dataset is to integrate a comprehensive
one by capturing bio-entities, disambiguated authors(distinguish between
different authors who may share the same or similar names), funding, and
affiliation information from the literature. Using this exact idea, A
knowledge graph called "PubMed Knowledge Graph" (**PKG**) is generated
by extracting bio-entities from 29 million PubMed abstracts,
disambiguating author names, integrating funding data through the
National Institutes of Health (NIH) ExPORTER, collecting affiliation
history and educational background of authors from ORCID, and
identifying fine-grained affiliation data from MapAffil.

<figure>
<img
src="https://cdn-images-1.medium.com/max/1024/0*61EpcIk903TKslvl" />
<figcaption>Bio-entity integration framework for PKG. First,
Bidirectional Encoder Representations from Transformers for Biomedical
Text Mining (BioBERT) is applied to extract bio-entities from 29 million
PubMed abstracts. Then, two existing high-quality author disambiguation
datasets, Author-ity and Semantic Scholar, are integrated. Eventually,
additional fields from credible sources are integrated into the dataset.
(Jian Xu et al., 2018) Article Link: <a
href="https://arxiv.org/abs/2005.04308">https://arxiv.org/abs/2005.04308</a></figcaption>
</figure>

Being an open dataset, PKG contains rich information ready to be
deployed, facilitating the effortless development of applications such
as finding experts, searching bio-entities, analysing scholarly impacts,
and profiling scientist's careers.

### Bio-LDA

Statistical modelling techniques such as Latent Dirichlet Allocation
(LDA) have been used nowadays to make the automated identification of
topics from large document collections and corpora possible. LDA model
is a model for a text corpus viewed as a collection of bags of words. It
assumes that people write an article with several topics in mind; each
topic is associated with a different conditional distribution over a
fixed set of words. A collection of documents, in this case, can be seen
as being generated by the same set of topics with different probability
distributions for each document. While previous applications of LDA in
the biomedical domain have yielded several benefits, few considered the
extension of the LDA model to include bio-terms.

A Bio-LDA model was proposed recently, extending the LDA model by
incorporating bio-terms as input variables to the classic LDA model. To
train the model, a great number of bio-terms are extracted through the
biomedical journals and papers provided on the PubMed database.

Experimental results show that Bio-LDA, in contrast to natural language
processing methods, can automatically derive a collection of topics of
related biological terms that map to clearly understandable biological
themes. The topics created using Bio-LDA are also surprisingly succinct
in identifying the bioterms associated with particular topic areas.

#### Conclusion

PubMed, as the primary tool for electronically searching and retrieving
biomedical literature, can be used either in a more intelligent way or
to make contributions to other fields of research. In this post, we
discussed three different directions about the usage of PubMed.
Moreover, we have also shared some promising scopes that could be used
for future exploration of these approaches. Through continuous
investigation and refinement, we believe that the use of PubMed can open
up exciting opportunities for us in the future.

![](https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c82fe27a0cad){width="1"
height="1"}


An introduction to tools supporting PubMed Author:  Xuzeng He (
<strong>
 ORCID:
</strong>
0009–0005–7317–7426)  PubMed is a free Web literature search service developed and maintained by the National Center for Biotechnology Information (
<strong>
 NCBI
</strong>
), and it is also a part of NCBI’s Entrez retrieval system that provides access to a diverse set of 38 databases.

Rogue Scholar Posts

Enhancing PubMed: from a Medical Database to beyond