6.1 Cognitive Computing
What is cognitive computing? IBM’s Watson is well known for cognitive computing. What is the difference between cognitive computing and AI? They could be
confused but are not similar. In general, while AI mimics the human brain, cognitive
computing aims to complement human beings by providing advice to make better
decisions and reinforcing human abilities. Such ability is called augmented
intelligence.
The word cognitive has the meaning “based on empirical knowledge.” Cognitive
computing is a system in which a computer not only processes instructions given by
humans but also thinks and learns as if it were a human and presents materials that
support human decision-making. Cognitive computing can understand not only
numbers and simple text but also unstructured data such as natural language, images,
sounds, and human expressions. Approximately 80% of the data in the world are said
to be unstructured data. Augmented intelligence makes it possible to understand and
process much more complex data than ever before. The University of Tokyo
Medical Research introduced IBM Watson Genomic Analytics (WGA) in July
2015. It was trained at the New York Genome Center at the time of introduction;
subsequently, it read over 20 million Medline data (literature abstract), over 15 million patent data, COSMIC (Catalog of Somatic Mutations in Cancer, UK), ClinVar
(Genomic Variation and Information on health, NIH USA), and National Cancer
Institute Pathways (NIH USA).
Systems prior to third-generation computing, such as cognitive computing, have
mostly dealt with problems that have only one solution; however, cognitive computing addresses questions that have multiple solutions. It has the ability to find the
best answer or to answer vague questions. In addition, by receiving feedback on the
answers that are derived, it has a deep learning function that allows it to learn on its
own and evolve so that accurate answers can be provided.
6.2 Panel Analysis and Whole Genome Sequence Analysis
IBM Watson Genomic Analytics has been introduced as whole genome sequencing
(WGS). Laskin et al. (2015) compared the detection rates of actionable mutations
with panel analysis and WGS in the terminal cancer patient group in detail. When
WGS was used for the terminal cancer patient group, actionable mutations were
found in 55 out of 78 patients, and 23 of them were actually treated. By contrast,
when the panel analysis was used for the same patient group, 73% of 81 patients had
mutations that were informative but not actionable (55% of those mutations were
TP53). In addition, 23% of the patients were not found in the panel. The panel
reported that nothing was eventually found to lead to an action (Laskin et al., 2015).
Although panel analysis is continuously innovating, as the above-mentioned
article points out, whole genome sequencing can be superior if it can be performed
20
H. Matsushita
What is cognitive computing? IBM’s Watson is well known for cognitive computing. What is the difference between cognitive computing and AI? They could be
confused but are not similar. In general, while AI mimics the human brain, cognitive
computing aims to complement human beings by providing advice to make better
decisions and reinforcing human abilities. Such ability is called augmented
intelligence.
The word cognitive has the meaning “based on empirical knowledge.” Cognitive
computing is a system in which a computer not only processes instructions given by
humans but also thinks and learns as if it were a human and presents materials that
support human decision-making. Cognitive computing can understand not only
numbers and simple text but also unstructured data such as natural language, images,
sounds, and human expressions. Approximately 80% of the data in the world are said
to be unstructured data. Augmented intelligence makes it possible to understand and
process much more complex data than ever before. The University of Tokyo
Medical Research introduced IBM Watson Genomic Analytics (WGA) in July
2015. It was trained at the New York Genome Center at the time of introduction;
subsequently, it read over 20 million Medline data (literature abstract), over 15 million patent data, COSMIC (Catalog of Somatic Mutations in Cancer, UK), ClinVar
(Genomic Variation and Information on health, NIH USA), and National Cancer
Institute Pathways (NIH USA).
Systems prior to third-generation computing, such as cognitive computing, have
mostly dealt with problems that have only one solution; however, cognitive computing addresses questions that have multiple solutions. It has the ability to find the
best answer or to answer vague questions. In addition, by receiving feedback on the
answers that are derived, it has a deep learning function that allows it to learn on its
own and evolve so that accurate answers can be provided.
6.2 Panel Analysis and Whole Genome Sequence Analysis
IBM Watson Genomic Analytics has been introduced as whole genome sequencing
(WGS). Laskin et al. (2015) compared the detection rates of actionable mutations
with panel analysis and WGS in the terminal cancer patient group in detail. When
WGS was used for the terminal cancer patient group, actionable mutations were
found in 55 out of 78 patients, and 23 of them were actually treated. By contrast,
when the panel analysis was used for the same patient group, 73% of 81 patients had
mutations that were informative but not actionable (55% of those mutations were
TP53). In addition, 23% of the patients were not found in the panel. The panel
reported that nothing was eventually found to lead to an action (Laskin et al., 2015).
Although panel analysis is continuously innovating, as the above-mentioned
article points out, whole genome sequencing can be superior if it can be performed
20
H. Matsushita