Creating an ignorance-base: Exploring known unknowns in the scientific literature

Abstract

Background: Scientific discovery progresses by exploring new and uncharted territory. More specifically, it advances by a process of transforming unknown unknowns first into known unknowns, and then into knowns. Over the last few decades, researchers have developed many knowledge bases to capture and connect the knowns, which has enabled topic exploration and contextualization of experimental results. But recognizing the unknowns is also critical for finding the most pertinent questions and their answers. Prior work on known unknowns has sought to understand them, annotate them, and automate their identification. However, no knowledge-bases yet exist to capture these unknowns, and little work has focused on how scientists might use them to trace a given topic or experimental result in search of open questions and new avenues for exploration. We show here that a knowledge base of unknowns can be connected to ontologically grounded biomedical knowledge to accelerate research in the field of prenatal nutrition.

Results: We present the first ignorance-base, a knowledge-base created by combining classifiers to recognize ignorance statements (statements of missing or incomplete knowledge that imply a goal for knowledge) and biomedical concepts over the prenatal nutrition literature. This knowledge-base places biomedical concepts mentioned in the literature in context with the ignorance statements authors have made about them. Using our system, researchers interested in the topic of vitamin D and prenatal health were able to uncover three new avenues for exploration (immune system, respiratory system, and brain development) by searching for concepts enriched in ignorance statements. These were buried among the many standard enriched concepts. Additionally, we used the ignorance-base to enrich concepts connected to a gene list associated with vitamin D and spontaneous preterm birth and found an emerging topic of study (brain development) in an implied field (neuroscience). The researchers could look to the field of neuroscience for potential answers to the ignorance statements.

Conclusion: Our goal is to help students, researchers, funders, and publishers better understand the state of our collective scientific ignorance (known unknowns) in order to help accelerate research through the continued illumination of and focus on the known unknowns and their respective goals for scientific knowledge.

Keywords: Epistemology; Information extraction; Knowledge representation; Knowledge-base; Natural language processing.

Figure 1:

Relationship between society, maternal nutrition (vitamin D), and the effects on mother and offspring: a Sankey diagram created based on Figure 3 from [10]. The orange color represents the findings from the exploration methods that the concepts related to brain development and immune system are enriched in ignorance statements and possible novel avenues to explore. SES/SDC = socioeconomic status/sociodemographic characteristics; BP = blood pressure; GDM = gestational diabetes mellitus.

Figure 2:

Network representation of the ignorance-base: The top right corner is the literature connecting the articles via segmented sentences (in blue) to the ignorance taxonomy (in yellow) through the ignorance classifiers (the annotated lexical cues). The sentences also connect to the biomedical concepts on the left with PheKnowLator [26, 27] using the biomedical concept classifiers with the ontologies of interest in bold and larger font.

Figure 3:

Ignorance vs. Standard Approach Results Chart: The interpretation of the results comparing the ignorance approach to the standard approach.

Figure 4:

Exploration by Experimental Results (gene list) pipeline: The results are in yellow highlights for the example presented here. For exploration at the end of the pipeline, the three not highlighted are the same as exploration by topic and the three highlighted are the new additions based on a gene list.

Figure 5:

Summary information for the ignorance-base. The ignorance-base is a combination of biomedical concept classifiers and ignorance classifiers over a corpus of prenatal nutrition articles. The network representation connected the literature to the ignorance theory and biomedical concepts via PheKnowLator [26, 27].

Figure 6:

Article date distribution for the ignorance-base (1939–2018).

Figure 7:

Ignorance taxonomy embedded in the research context: Starting from the top, research starts from known unknowns or ignorance. Our ignorance taxonomy is in green (an ignorance statement is an indication of each ignorance category) with knowledge goals underneath. Research is then conducted based on the knowledge goals to get answers; these then filter back to the known unknowns to identify the next research questions.

Figure 8:

Exploring the ignorance-base by Vitamin D: Searching the ignorance-base for vitamin D yielded many articles and sentences that can be explored using ignorance statements to find new research questions, including immune system and brain development.

Figure 9:

Term frequency results: Frequent Biomedical Concepts and Words in (a) ignorance approach vitamin D ignorance statements and (b) standard literature approach vitamin D sentences. Word clouds using words and biomedical concepts are on the right and left respectively. Also underneath are frequency tables of the top 5 most frequent concepts or words.

Figure 10:

Comparison of standard and ignorance enrichment: A Venn diagram of biomedical concept enrichment between just vitamin D (pink) and ignorance vitamin D (green) sentences. Next to each bubble are concepts in their respective enrichment orders. The concepts in the middle are the overlap and the numbers correspond to the enrichment position for the ignorance vitamin D enrichment, with the overlap position in parentheses. Skeleton of manus is an error and is actually annotating autoimmune as in the parentheses. *Statistically significant with FDR but not family-wise error.

Figure 11:

Ignorance-category enrichment: Ignorance vitamin D sentences compared to all ignorance sentences. The 10 categories highlighted in green were enriched.

Figure 12:

How ignorance changes over time: A bubble plot of vitamin D and immune system sentences (including non-ignorance sentences). The x-axis is the articles sorted by time. The y-axis is the ignorance categories. Each bubble represents the portion of sentences in each article in that ignorance category (scaled by the amount of total ignorance sentences in the category). For example, future prediction only appears in two different articles and is basically split in half between both.

Figure 13:

Enhancing canonical enrichment analysis using the ignorance-base: DAVID enrichment analysis for the gene ontology (GO) in relation to the ignorance-base. The DAVID initial analysis is on the left with 42 of the 43 genes found in DAVID mapping to 159 GO concepts. The right is a breakdown of where the 51 enriched GO concepts from DAVID fall within the ignorance-base.

Figure 14:

Comparison of DAVID and ignorance enrichment: A Venn diagram of gene ontology enrichment between DAVID (pink) and the ignorance-base (green). In parentheses are the total number of concepts found in each category without enrichment. Next to each bubble are the top three concepts for each enrichment method. The concepts in the middle are the overlap. *Statistically significant with FDR but not family-wise error.