Information Technology Reference
In-Depth Information
5.3.1
Imputing Wisdom from Available Data
The tsunami of data that are generated across the spectrum of biomedicine underlie
the belief that the wisdom of the masses will lead to a revolution in biomedical
research and offer unprecedented improvements in quality of patient care [ 44 - 46 ].
Advances in computation, both from effi ciency and algorithmic innovations, have
even suggested that machine-based knowledge extraction systems will help meet
the demand for increased health care professionals. It is important to temper these
types of beliefs with the reality of how data are currently available. Indeed, the
computational advances that have arose in the last 20 years have led to unprece-
dented ability to analyze more data than ever believed possible. However, the vol-
ume at which data are being generated greatly exceeds the potential to leverage
them in a meaningful and, perhaps more importantly in the context of healthcare,
timely manner.
The use of bibliome mining techniques are thus needed for two critical and inter-
dependent activities: (1) to partition the growing landscape of biomedical data into
relevant versus irrelevant for a given context; and, (2) from within relevant corpora,
identify associations that either by themselves or in combination with other data
give useful insights that may have not been otherwise obvious. These activities
should both be driven by the common goal to identify the most relevant data, at the
most relevant time, and in the most effi cient manner. The general challenges that
characterize the diffi culty in leveraging big data (described in Chap. 7 ) are general-
izable to data described within a collection of biomedical literature.
The supporting biomedical infrastructure thus requires a repository of knowl-
edge that can underpin future innovations. Especially in the era of big data and high
throughput data generation, it is essential to be able to identify meaningful knowl-
edge from just highly occurring concepts. And, herein lies the greatest challenge:
deciphering useful knowledge from simply frequently occurring correlations that
are a result of aberrations in how the data are generated or recorded. In order for
knowledge to be considered wisdom, signifi cant curation effort is required. No mat-
ter how much data are generated, or how much purported knowledge is imputed, the
amount of usable storage of wisdom will always be fi nite. This has signifi cant impli-
cations. It is therefore important to appreciate that not all knowledge need be cata-
logued as wisdom, and yet still be accepted as useful for a given scenario.
Mining the bibliome thus requires a realistic understanding of what problem is
aiming to be solved. The needs for bench scientists, clinical practitioners, and com-
munity participants may be met by a common set of biomedical literature; however,
the detailed approach for mining the appropriate level of knowledge will undoubt-
edly require differing computational or algorithmic approaches. For example, a
bench scientist may be satisfi ed with the generation of testable hypotheses, whereas
a clinical practitioner would require some statistical support for a proposed hypoth-
esis, or whereas a community participant would require some understanding of
broader implications of a hypothesis. It is essential to refl ect that bibliome mining
may offer some glimpse at new knowledge, but many mining exercising will only
Search WWH ::




Custom Search