INSIGHTS | PERSPECTIVES
1536 23 DECEMBER 2016 • VOL 354 ISSUE 6319 sciencemag.org SCIENCE
By N. Dello Russo
In the past many humans regarded the appearance of a comet as an omen or harbinger of doom, but we now know them as time capsules from the forma- tion of the solar system. Created from the gradual accretion of rocks, dust,
and ice, comets have been preserved in the
“deep freeze” of space far from the Sun,
likely retaining in the present much of their
formative character. The present-day struc-
ture and composition of comets provide a
window into the conditions that prevailed
during the birth of our solar system. Comet
67P/Churyumov-Gerasimenko, the target of
the Rosetta mission, is a member of the Ju-
piter-family dynamical group of comets that
have likely spent most of the last few billion
years stored in the ring-shaped disc of icy
bodies beyond the orbit of Neptune known
as the Kuiper Belt. Subtle gravitational in-
teractions can eventually push icy bodies
in the Kuiper Belt reservoir toward the Sun
where, under the gravitational influence of
Jupiter, they evolve into short-period or-
bits around the Sun. Once its journey into
the inner solar system begins, however, its
More than a day in the
life of a comet
Learning how a comet evolves can reveal
the nature of the early solar system
Johns Hopkins Applied Physics Laboratory,
11100 Johns Hopkins Road, Laurel, MD 20723, USA.
or not) can be reconstructed from structured
billing codes that indicate assigned diagnoses
and performed procedures in the medical record. Much of the richness of the EHR, however, resides in the free text of the encounter
notes and reports, and emerging natural
language processing platforms should allow
more sophisticated extraction and curation
of this information.
Even with the advantages of coupling
dense rare variant genomic data with EHR-based phenotyping, there are several caveats
that must be considered. The absence in the
EHR of a phenotype for any given individual
does not mean that the participant lacks the
trait; it may have never been tested for, or
the participant may have sought care outside of the health system captured by that
EHR. In addition, subject enrollment in an
EHR cohort reflects the underlying catchment of the health system. In this regard,
it is important to note that the population
in DiscovEHR was >98% white and only
~1% African American. Moving forward,
it will be important to construct cohorts
with greater racial and ethnic diversity, as
has been accomplished in the initial efforts
of the U.S. Department of Veterans Affairs’
Million Veteran Program (10) and is aspired
to in the proposed U.S. National Institutes
of Health’s “All of Us” cohort (11, 12).
Continued advances and price reductions in next-generation sequencing will
make WES, and ultimately whole-genome
sequencing, feasible on an increasingly
larger scale. Better annotation of rare coding (and noncoding) variants and new statistical approaches will be critical to the
analyses of these large data sets. Efforts
to apply natural language processing and
machine learning to EHR data will reveal
substantially improved phenotypes and
subphenotypes for analysis. These endeavors will require new experimental, computational, and statistical techniques to
maximize biological discovery. With the appropriate investment, these undertakings
will provide new insights into the genomic
basis of health and disease and pave the
way for the application of precision medicine on a population health scale. j
1. F.E.Dewey et al., Science 354,aaf6814(2016).
2. N.S.Abul-Husn et al., Science 354,aaf7000(2016).
3. O.Gottesman et al., Genet. Med.15,761(2013).
4. R. Do et al ., Nature 518, 102 (2015).
5. D. G. Mac Arthur et al ., Nature 508, 469 (2014).
6. A. R. Majithia et al ., Nat. Genet. 48, 1570 (2016).
7. J. C. Denny etal ., Nat.Biotechnol. 31, 1102 (2013).
8. S. S. Gidding et al ., Circulation 132, 2167 (2015).
9. E. M. deGoma et al., Circ. Cardiovasc. Genet .9, 240 (2016).
10. J. M. Gaziano et al ., J. Clin. Epidemiol. 70, 214 (2016).
11. F.S.Collins, H.Varmus, N. Engl. J. Med. 372,793(2015).