Nature Reviews Drug Discovery | AOP, published online 18 July 2014; doi:10.

1038/nrd4336

ANALYSIS
The discovery of first‑in‑class drugs:
origins and evolution
Jörg Eder, Richard Sedrani and Christian Wiesmann
Abstract | Analysis of the origins of new drugs approved by the US Food and Drug
Administration (FDA) from 1999 to 2008 suggested that phenotypic screening strategies
had been more productive than target-based approaches in the discovery of first‑in‑class
small-molecule drugs. However, given the relatively recent introduction of target-based
approaches in the context of the long time frames of drug development, their full impact
might not yet have become apparent. Here, we present an analysis of the origins of all
113 first‑in‑class drugs approved by the FDA from 1999 to 2013, which shows that the
majority (78) were discovered through target-based approaches (45 small-molecule drugs
and 33 biologics). In addition, of 33 drugs identified in the absence of a target hypothesis,
25 were found through a chemocentric approach in which compounds with known
pharmacology served as the starting point, with only eight coming from what we define
here as phenotypic screening: testing a large number of compounds in a target-agnostic
assay that monitors phenotypic changes. We also discuss the implications for drug
discovery strategies, including viewing phenotypic screening as a novel discipline rather
than as a neoclassical approach.

First‑in‑class drugs
Since Drews and Ryser1 published an analysis on escalating An excellent recent analysis of new medicines that
Drugs that modulate an as-yet research and development (R&D) costs in the pharma­ were approved by the FDA during the 10‑year period
unprecedented drug target or ceutical industry and discussed the consequences if this from 1999 to 2008 found that, of the first‑in‑class drugs
biological pathway. increase continued, many additional studies and per­ that are small molecules, 28 were discovered through
spectives have been published on the trends, metrics phenotypic screening, whereas 17 originated from target-
Phenotypic screening
The testing of a large number and measures that drive R&D productivity. Several have based approaches12. Given that this was an era in which
of — in most cases randomly offered possible solutions to increase R&D efficiency 2–8, the focus and investment was heavily biased towards
selected — compounds in a but some are rather pessimistic and raise doubts about target-based approaches, the apparent greater success
systems-based assay.
the sustainability of the current drug discovery model9–11. of phenotypic screening in the discovery of innovative
Target-based approaches
In fact, the numbers are concerning. Over the past six small-molecule drugs raises several important questions.
Hypothesis-based approaches decades the average inflation-adjusted cost of bringing Has the pharmaceutical industry invested in the wrong
that aim to manipulate a a new drug to market has been increasing constantly and technologies for almost three decades? Should the indus­
biological system by pharma­ is doubling approximately every 9 years, despite scien­ try return to a ‘classical’ phenotypic approach to drug
co­­­log­ically modulating a specific
tific discoveries and technological advances that include discovery, as postulated by some13–16? What is the basis
component or target (an
enzyme, receptor, and so on). modern molecular biology methods, high-throughput for the apparent superiority of phenotypic screening over
screening, structure-based drug design, combinatorial target-based approaches, and what are the implications
and parallel chemistry, and the sequencing of the human for future drug discovery projects?
Novartis Institutes for
BioMedical Research, genome8. These innovations have allowed a rational,
Fabrikstrasse 16, target- and hypothesis-driven approach to drug discov­ A definition of phenotypic screening
CH‑4002 Basel, Switzerland. ery and were implemented with the promise of greatly To try to answer these questions, we analysed the origins
Correspondence to J.E. enhancing the productivity of R&D. However, so far of the first-in-class drugs approved by the FDA between
e‑mail:
[email protected]
there has been little apparent impact of these advances 1999 and 2013, an extension of 5 years over the previous
doi:10.1038/nrd4336 on the number of drug approvals by the US Food and analysis12. We considered this extension to be important
Published online 18 July 2014 Drug Administration (FDA). in evaluating the impact of target-based approaches,

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 1

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

a First-in-class drugs systems-based approach using a target-agnostic assay

that monitors phenotypic changes. We believe that this
is how the term is understood and used in most research
Target-based Systems-based Other laboratories today.
Using this definition, it is apparent that not all systems-
based approaches to drug discovery rely on phenotypic
Small-molecule Chemocentric Phenotypic
(NDA)
Biologic (BLA)
approaches screening screening. Aspirin, for example, was not discovered
through phenotypic screening in this sense, but through
• Screening the isolation and further derivatization of an active
• Chemocentric
• Chemocentric/ ingredient from a plant extract, the pharmacological
rational design activity of which was known for hundreds of years18. In
fact, before the invention of modern molecular biology
tools in the mid‑1980s and the technological advances
b 50 c 16
in high-throughput screening in the early 1990s, most
14 drugs were discovered based on studies with a particular
40
12 compound or compound class in a systems-based and
30 10 target-agnostic manner 19. For this approach, we would
Number

Number

45 like to coin the term ‘chemocentric drug discovery’ and

8
(41%)
20 33 6
propose to categorize drug discovery approaches as
(30%) 25 shown in FIG. 1a, with systems-based drug discovery sub­
(23%) 4
10 divided into two categories: phenotypic screening and
8 2
chemocentric approaches. The latter approaches typically
0
(7%) 0 include the identification of an active ingredient from a
Target-based Systems-based 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13
78 (70%) 33 (30%) 19 20 20 20 20 20 20 20 20 20 20 20 20 20 20 plant or microbial extract with known pharmacological
Small-molecule Total
activity (for example, aspirin) or the derivatization of a
Biologic Target-based pharmacologically active natural substance (for example,
Chemocentric Systems-based prostaglandins, steroids, nucleosides, amino acids and
Phenotypic screening
biogenic amines) or synthetic chemical, often based on
Figure 1 | Discovery of first‑in‑class drugs approved by the FDA between 1999 serendipitous findings made decades before. We found
and 2013. a | First‑in‑class drugs were classified according to whether they were many such cases in our analysis, and some are highlighted
Nature Reviews | Drug Discovery
discovered in a systems-based, target-agnostic manner or using a hypothesis-driven, in BOX 1.
target-based approach. Central to the discovery of systems-based drugs was either a With the advent of gene cloning and sophisticated
phenotypic screen or a chemocentric approach starting from a known compound or molecular biology techniques in the mid‑1980s, it was
compound class. Target-based drugs were categorized into small-molecule drugs possible to work in a more hypothesis-based, rational
and biologics depending on whether they were approved by the US Food and Drug and systematic manner on particular protein targets.
Administration (FDA) under a new drug application (NDA) or biologics license
Many pharmaceutical companies quickly switched to this
application (BLA), respectively. b | The majority of first‑in‑class drugs were discovered
through target-based approaches with slightly more small-molecule drugs than
new approach, which is typically referred to as target-
biologics. Most drugs that were discovered through systems-based approaches based drug discovery. High-throughput technologies
originated from a known compound or compound class (that is, a chemocentric for screening large compound libraries in target-based
approach), and only a few were based on a phenotypic screen as defined in this article. assays have been used to discover many new, synthetic
c | There is no statistically significant trend over the 15‑year period that would indicate or naturally occurring pharmacologically active com­
a growing superiority of one approach over the other. However, from 2003 onwards pounds with low molecular mass20. Molecular biology
(with two exceptions), the number of newly approved target-based first‑in‑class drugs techniques have also enabled the development of thera­
exceeds that of system-based drugs, and since 1999 the level of systems-based peutic biologics, such as monoclonal antibodies that are
drugs has been constantly low. specific for a particular protein target21. For our analysis,
we have therefore divided target-based drugs into these
two categories (FIG. 1a).
given their relatively recent introduction in the context
of the long time frames of drug discovery and develop­ Analysis
ment. We also realized that the conclusions that can be First‑in‑class drugs and their origins. Our analysis
Small-molecule drugs drawn from the data depend on a clear definition of the covers a time frame of 15 years (1999–2013), during
Drugs with a low molecular drug discovery approaches and on careful use of termi­ which 113 first‑in‑class drugs were approved by the FDA
mass (typically <1,000 Da); nology. The term ‘phenotypic screening’ in particular (see the Drugs@FDA database). The results are shown in
this includes synthetic
appears to be used rather loosely and with different FIGS 2,3,4 and the data are listed in Supplementary infor­
drugs, natural products
(or derivatives) and natural meanings. In the original analysis12, and also in a subse­ mation S1 (table). Drugs were designated as first‑in‑class
substances (or derivatives). quent paper focused on first‑in‑class drugs17, phenotypic drugs based on their modulation of an — until then —
screening was considered to encompass all non-target- unprecedented target or biological pathway. This was
Systems-based approach based approaches to drug discovery. However, for the considered to be independent of the mechanism of mod­
Hypothesis-agnostic assay
or approach that monitors or
purpose of this analysis we define phenotypic screening ulation; that is, if two drugs modulate the same target
is based on a phenotypic more specifically as the testing of a large number of with the same biological consequence but bind to dif­
change in vitro or in vivo. (in most cases randomly selected) compounds in a ferent sites (for example, the active site versus allosteric

2 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

Box 1 | Examples of chemocentric drug discovery

Ingenol angelate
Plants of the Euphorbiaceae family have been used for the treatment of cancers and warts since at least 400 bc.
The isolation of ingenol angelate as one of the active principles in 1983 led to the development of a drug that was
approved by the US Food and Drug Administration (FDA) in 2012 for the treatment of actinic keratosis55.
Nitisinone
Nitisinone, which was approved in 2002 for the treatment of hereditary tyrosinaemia type 1, is based on a compound
class developed as herbicides in the 1970s (originating from natural products, beta-triketones, which were known for
decades). The toxicological profile, together with the subsequently elucidated mechanism of action, led to the
therapeutic hypothesis and start of clinical development in 1989 (REF. 56).
Varenicline
Varenicline, another drug derived from a natural product, was approved in 2006 for smoking cessation and is based on
cytisine, which has been known since 1912 as a substance with nicotine-like activity57.
Nelarabine
Nelarabine is an example of a drug derived from a natural substance. It was known since the 1960s that some nucleoside
analogues (for example, cytosine arabinoside) have antitumour activity58. Later, deoxyguanosine analogues59 and
arabinosyl guanine60 were found to be selectively toxic for leukaemic T cells. Nelarabine, which was approved in 2005 for
the treatment of T cell acute lymphoblastic leukaemia and T cell lymphoblastic lymphoma, is a water-soluble prodrug of
arabinosyl guanine61.
Docosanol
Docosanol is a synthetic low-molecular-mass molecule. Based on the finding that butylated hydroxyl toluene, a food
additive, integrates into membranes and disturbs them, the antiviral activity of long-chain unsaturated monoglycerides
and alcohols was discovered, and docosanol was thus approved as antiviral agent in 2000 (REF. 62).
Ezogabine
Ezogabine, also known as retigabine, is another synthetic low-molecular-mass molecule and was synthesized as a
back‑up compound to flupirtine, which was discovered in the 1960s in an effort to find novel analgesics with a mode
of action different from opiates. Ezogabine is a 2,3,6‑triaminopyridine derived from the known analgesic pyridium
(discovered in the 1930s). Later, ezogabine as well as flupirtine were found to possess anti-epileptic activity in addition
to their analgesic activity in various animal models, and ezogabine was approved for the treatment of epilepsy in 2011
(REF. 63).

Chemocentric approaches
Drug discovery approaches
based around a specific site of an enzyme), only the drug that was approved first endothelial growth factor (VEGF) in 1983 (REF. 25) and
compound or compound class. has been categorized as first-in-class. For the purpose its cloning in 1989 (REF. 26) facilitated the discovery of
Chemocentric approaches have of our analysis, we excluded diagnostic drugs such as bevacizumab, the first VEGF-specific antibody 27. For
made a substantial contribution
contrast agents. our analysis, we have chosen the purification of VEGF
both to drugs originating from
systems-based approaches We searched scientific publications and the patent as the starting point of drug discovery efforts. Another
and to drugs originating from literature using the chemical structure of the molecule example is the discovery of imatinib for the treatment
target-based approaches. and its mechanism of action to identify the following: the of chronic myelogenous leukaemia (CML)28. A chro­
origins of the relevant chemotype; the findings that led mosomal abnormality, the Philadelphia chromosome,
Natural substance
(or derivative)
to the formulation of the therapeutic hypothesis and the was discovered in 1960 in white blood cells of patients
A chemical substance link to the final indication; the methods and technolo­ with CML29. In 1973 the Philadelphia chromosome was
(or derivative thereof) gies that were used for the discovery of the drug, and shown to be a translocation between chromosomes 9 and
produced by a living organism the first publication on the final drug molecule. We have 22 (REF. 30). A series of subsequent discoveries resulted
found in nature that usually
defined the starting point for drug discovery as the pub­ (in 1985) in the insight that the chromosomal translo­
has pharmacological or
biological activity. For this lication of the key finding that, at the time, enabled the cation leads to the expression of the BCR–ABL fusion
article we arbitrarily excluded initiation of dedicated drug discovery efforts. In many protein and led to the hypothesis that its tyrosine kinase
natural products from natural cases, these key initial findings were the identification activity drives malignant transformation31. Imatinib was
substances to keep the of the target or chemotype. In our analysis, we have not subsequently developed as an inhibitor of the BCR–ABL
former as a separate class
of compounds.
taken into account the fundamental research that led kinase. Given the scope of this analysis, we selected the
to these key findings, which in itself often constituted a discovery of the BCR–ABL fusion protein as the starting
Biologics series of important discoveries. Therefore, our definition point for drug discovery.
Defined here as all drugs of the starting point arbitrarily separates the foundation of We also realize that what we have identified as a first
approved under a biologics
scientific studies from drug discovery; however, in reality publication might not always represent the exact starting
license application (BLA)
by the US Food and Drug they are closely linked. point of discovery efforts towards a particular drug. This
Administration (FDA); usually For example, the capacity of tumour cells to stimu­ might pertain more to systems-based drugs, as the first
antibodies and other proteins. late angiogenesis was discovered in 1945 (REF. 22) and the publication of the chemotype — which usually is also the
presence of soluble tumour-derived factors was demonstr­ first publication of such projects — may have occurred
Chemotype
A family of molecules that
ated in 1968 (REF. 23). This led to the formulation of the several years after the initiation of drug discovery activi­
possess the same core ‘anti-angiogenesis’ therapeutic concept for the treatment ties. In addition, publication intensity was substantially
structure or scaffold. of tumours24. The subsequent purification of vascular lower before the 1980s. For target-based drugs, however,

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 3

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

Year 1950 1960 1970 1980 1990 2000

Cilostazol
Denileukin diftitox
Orlistat
Zanamivir
Sirolimus
Pemirolast
Levetiracetam
Aminolevulinic acid
Zonisamide
Gemtuzumab (withdrawn in 2010)
Docosanol
Mifepristone
Caspofungin
Alemtuzumab
Imatinib
Anakinra
Bosentan
Drotrecogin alfa (withdrawn in 2011)
Nitisinone
Ezetimibe
Nitazoxanide
Rasburicase
Alefacept
Enfuvirtide
Pegvisomant
Aprepitant
Agalsidase beta
Laronidase
Patent
Chemotype or first antibody Gefitinib
Final molecule
Target Bortezomib
Mechanism of action
Therapeutic concept Omalizumab
Indication
Protein, enzyme or target cloned Miglustat
Traditional medicine Daptomycin
Memantine
Efalizumab (withdrawn in 2009)

Figure 2 | Chronology of the discovery of first‑in‑class drugs approved between 1999 and 2003. All 113
first‑in‑class drugs approved by the US Food and Drug Administration (FDA) between 1999 and 2013 are listed in the
Nature
order of their approval date in FIGS 2,3,4. This figure shows those approved in the 5-year periodReviews | Drug
from 1999 Discovery
to 2003.
The colour coding for the different approaches, systems-based phenotypic screening, systems-based chemocentric,
target-based small-molecule drug and target-based biologic, is defined as in FIG. 1. Other drugs are represented by grey
arrows. Important points in the discovery chronology of each drug, such as the publication year of the patent covering
the final drug molecule and the publication year of the final molecule, are indicated on the arrows using the symbols in the
key. The first publication, usually the identification of the chemotype, target or concept, does not necessarily mark
the exact starting point of discovery efforts towards a particular drug; this is symbolized by the fading at the beginning
of each arrow. For details of each drug, see Supplementary information S1 (table).

4 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

we believe that many projects were initiated around the Drug types and discovery approaches. According to
time of the publication of the target hypothesis, which we our analysis and definitions, of the 113 first‑in‑class
take as the typical starting point of such projects. In fact, drugs, 33 (30%) were discovered through systems-
some target-based drug discovery projects might even based approaches and 78 (70%) were discovered from
have started later than that. Despite these uncertainties, target-based approaches (FIG. 1b); the numbers discov­
we think that our analysis gives a clear picture of the vari­ ered per year from each approach are shown in FIG. 1c.
ous drug discovery approaches and chronology of events. Two drugs were classified as ‘other’: aminolevulinic

Year 1960 1970 1980 1990 2000 2010

Cetuximab
Bevacizumab
Cinacalcet
Azacitidine
Acamprosate
Natalizumab
Palifermin
Ziconotide
Pramlintide
Exenatide
Galsulfase
Ramelteon
Nelarabine
Sorafenib
Abatacept
Conivaptan
Sunitinib
Ranolazine
Lubiprostone
Alglucosidase alfa

1912 Varenicline
Idursulfase
Vorinostat
Sitagliptin

Ca. 200 BC Sinecatechins

1939 Aliskiren
Eculizumab

Patent Maraviroc
Chemotype or first antibody
Final molecule Raltegravir
Target Tetrahydrobiopterin
Mechanism of action
Therapeutic concept Rufinamide
Indication
Protein, enzyme or target cloned Romiplostim
Traditional medicine

Figure 3 | Chronology of the discovery of first‑in‑class drugs approved between 2004 and 2008. The colour
coding for the different approaches, systems-based phenotypic screening, systems-based chemocentric, target-based
small-molecule drug and target-based biologic, is defined as in FIG. 1. Important points inNature Reviews | Drug Discovery
the discovery chronology
of each drug are indicated on the arrows using the symbols in the key. For details of each drug, see Supplementary
information S1 (table).

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 5

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

Year 1960 1970 1980 1990 2000 2010

Canakinumab
Artemether–lumefantrine
Ustekinumab
Ecallantide
Tocilizumab
Collagenase Clostridium
histolyticum
Carglumic acid
Denosumab
Fingolimod
Tesamorelin
Eribulin
Roflumilast
Belimumab
Ipilimumab
Abiraterone
Boceprevir and telaprevir
Rivaroxaban
Patent
Chemotype or first antibody Ezogabine
Final molecule
Target Vemurafenib
Mechanism of action Brentuximab vedotin
Therapeutic concept
Indication Icatibant
Protein, enzyme or target cloned
Traditional medicine Crizotinib
Ruxolitinib
Glucarpidase

Ca. 400 BC Ingenol mebutate

Vismodegib
Ivacaftor
Lucinactant
Mirabegron
Linaclotide
Ocriplasmin
Perampanel
Omacetaxine
Tofacitinib
Raxibacumab
Teduglutide
Lomitapide
Bedaquiline
Crofelemer
Mipomersen sodium
Dimethyl fumarate
Canagliflozin
Trametinib
Riociguat
Ibrutinib
Sofosbuvir

6 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc

Nature Reviews | Drug Discovery
© 2014 Macmillan Publishers Limited. All rights reserved
A N A LY S I S

Figure 4 | Chronology of the discovery of first‑in‑class drugs approved between 25 (33%) were biologics and five (7%) came from other
▶
2009 and 2013. The colour coding for the different approaches, systems-based approaches12. This discrepancy occurs for two reasons.
phenotypic screening, systems-based chemocentric, target-based small-molecule First, we consider biologics to be target-based drugs, as
drug and target-based biologic, is defined as in FIG. 1. Other drugs are represented there is little philosophical distinction in the hypothesis-
by grey arrows. Important points in the discovery chronology of each drug are
driven approach to drug discovery for small-molecule
indicated on the arrows using the symbols in the key. For details of each drug,
see Supplementary information S1 (table).
drugs versus biologics. Second, the past 5 years of our
analysis time frame have seen a significant increase in
the approval of first‑in‑class drugs, most of which were
discovered in a target-based fashion.
acid (a precursor of protoporphyrin, which is used for With regard to the second reason, it is interesting to
photodynamic therapy) and lucinactant (a peptide that look at the time frame for drug discovery projects. For
lowers alveolar surface tension, used for the treatment of all the drugs in our data set, we calculated the apparent
respiratory distress syndrome). median time from the first publication of the therapeu­
Target-based drugs are divided into 45 (41%) small- tic concept, target or chemotype to FDA approval, and
molecule drugs and 33 (30%) biologics. Biologics were found that it was 22 years. There was also a statistically
typically found by screening (antibodies) or rational significant difference in the median time frame for drugs
design. The starting points for target-based small-molecule that were discovered through systems-based versus
drugs were derived as follows: 21 from various screening target-based approaches: 25 and 20 years, respectively
methods (18 from high-throughput screening, one from (Supplementary information S2 (box)). Therefore, taking
fragment-based screening, one from in silico screen­ into account the fact that the tools needed to efficiently
ing and one from low-throughput screening); 18 from discover drugs in a hypothesis-driven manner — includ­
chemocentric approaches (for example, the starting points ing modern gene cloning and expression methods,
were analogues of known ligands for the target or related high-throughput screening, crystallography and the
targets); and six from rational design, in most cases based sequencing of the human genome — have only become
on a known substrate. established or sufficiently advanced between 1985 and
Of the 33 systems-based drugs, 25 were small-molecule 2000, it is not surprising that the impact of these tools
drugs that were discovered through chemocentric on target-based drug discovery may only have begun to
approaches (which were considered as a type of pheno­ become apparent in more recent years12.
typic screening in the previous analysis) 12. A further Interestingly, all but four systems-based drugs have
seven small-molecule drugs were discovered by pheno­ their beginnings before 1985, and one of those four
typic screening according to the more specific definition drugs — ezetimibe — originated from a target-based
we use: sirolimus, daptomycin, artemether–lumefantrine, drug discovery project until it was noticed that the lead
fingolimod, eribulin, bedaquiline and trametinib. The molecule had in vivo activity independently of, or in
other drug we classified as being discovered through addition to, its target-based activity 32. So, the finding
a phenotypic screening approach was alemtuzumab, that a considerable number of systems-based drugs have
an antibody directed against CD52 that was discov­ been approved over the past 10 years is likely to be due in
ered by raising antibodies against human peripheral part to the length of time these projects took.
blood mononuclear cells without knowledge of the The data also suggest that target-based drug discovery
target. might have helped reduce the median time for drug
Three of the small-molecule drugs originating from discovery and development. Closer examination of the
phenotypic screening (sirolimus, fingolimod and eribu­ differences in median times between systems-based
lin) are natural product-derived compounds that were approaches and target-based approaches revealed that
found by analysing a discrete number of extracts for the 5‑year median difference in overall approval time is
specific biological activity, with some prior knowledge largely due to statistically significant differences in the
on similar extracts (see Supplementary information S1 period from patent publication to FDA approval, where
(table)). Interestingly, the original phenotypic activity target-based approaches (taking 8 years) took only half
of sirolimus and that of the fingolimod precursor myri­ the time as systems-based approaches (taking 16 years)
Natural product ocin were unrelated to their therapeutic activity, which (Supplementary information S2 (box)).
(or derivative) only became apparent during follow‑up studies. Of the The pharmaceutical industry has often been criticized
Secondary metabolites
(or derivatives thereof) that
eight drugs discovered through phenotypic screen­ for not being sufficiently innovative. We think that our
are extracted from tissues of ing, three are anti-infective, four are anti-proliferative analysis indicates otherwise and perhaps even suggests
plants, marine organisms or or cytotoxic molecules, and one — fingolimod — is an that the best is yet to come as, owing to the length of time
microorganism fermentation immunosuppressive drug. between project initiation and launch, new technologies
broths.
such as high-throughput screening and the sequencing
Pharmacophore Target- versus systems-based approaches of the human genome may only be starting to have a
The steric and electronic At first glance, the results of our analysis appear to sig­ major impact on drug approvals. Target-based drug dis­
features of a ligand that are nificantly deviate from the numbers previously pub­ covery, together with modern screening technologies,
necessary to ensure optimal lished for first‑in‑class drugs, which reported that of has also greatly broadened the scope of pharmacophores
interactions with a biological
target structure and to trigger
the 75 first‑in‑class drugs discovered between 1999 and available for medicinal chemistry 20, and increased the
(or to block) its biological 2008, 28 (37%) were discovered through phenotypic number of ‘tool compounds’ to use for investigating bio­
response. screening, 17 (23%) through target-based approaches, logical systems, potentially leading to new therapeutic

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 7

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

a Total
Systems-based Target-based
30% 42% 3% (1)
(34) (47) 12% 50% 42% 38%
(4) (17) (33) (30)

35%
15% (12)
(17) 3%
(2)
13% (14) 17% (13)

Small-molecule, synthetic Natural product (derived) Natural substance (derived) Biologic

b Total
8% (9) 13% Systems-based Target-based
(15)
9% (3) 15%
10% 6% 21%
19% (11) 26% (12)
(2) (16)
(21) (9)
12%
18%
(9)
(6)
8% 5%
(9) (4)
4%
(4)
7% 10%
7% (8) (8) 12%
(8) 15% (9)
1% (5)
(1) 3%
22% 1% (2)
23% (26) 26% (9) (17) (1)

GPCR Protease NHR Other receptor Other mechanism

Kinase Ion channel Other enzyme Cytokine Unknown

Figure 5 | Distribution of first‑in‑class drugs according to the molecule type and target family. a | The distribution
according to molecule type (that is, the source of the drug molecule) differs between systems- and target-based
approaches. There is a strong bias for the use of natural products for systems-basedNature
approaches over
Reviews target-based
| Drug Discovery
approaches, whereas biologics have only had a major impact in the target-based space. Note that artemether–
lumefantrine was counted twice as it contains two drugs. b | The distribution of drugs across target families.
G protein-coupled receptors (GPCRs), kinases, proteases and ion channels constitute the major target families.
Kinase and protease inhibitors are prominent in the list of target-based drugs, but under-represented or absent from
that of systems-based drugs. By contrast, the fraction of drugs targeting ion channels is significantly higher among
systems-based drugs. Note that memantine was counted twice as it modulates GPCRs as well as ion channels.
NHR, nuclear hormone receptor.

targets and/or chemical starting points for new drugs. drugs, mainly low-molecular-mass synthetic compounds
One question that may be asked, although a rather (FIG. 5a). Although natural substances appear to be the
academic one, is what course the industry would have basis for a similar proportion of the drugs discovered
taken in the absence of these innovations: that is, at what by target- and systems-based approaches (17% and
point in time would the industry have run out of chemo­ 12% respectively), natural products were the basis for
types as starting points for the classical chemocentric a much more substantial proportion of systems-based
approach? approaches. Of the 33 systems-based drugs, 12 (35%) are
derived from natural products, whereas only two (3%)
Molecules and target families of the 78 target-based drugs are derived from natural
Of the 113 first‑in‑class drugs, 34 (30%) are biologics, products: orlistat, which is based on the lipase inhibitor
14 (13%) are natural products or compounds derived lipstatin; and canagliflozin, which is based on phlorizin.
Low-molecular-mass thereof, 17 (15%) are natural substance-derived mole­ Of the 14 natural product-derived first‑in‑class drugs
synthetic drug
Low-molecular-mass drugs that
cules (including peptides, but arbitrarily excluding natu­ approved between 1999 and 2013, five were discovered
are not derived from natural ral products to keep this class of compounds separate by screening approaches (four by systems-based pheno­
products or natural substances. for the analysis) and 47 (42%) are other small-molecule typic screening and one by target-based high-throughput

8 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

screening) and nine were discovered by chemocentric Today, the pharmaceutical industry has largely taken
approaches. The latter includes four drugs that were a step back from this ‘brute force’ approach, realizing
found by ethnobotanical approaches through the isola­ that this seemed to instead hamper creativity, innovation
tion and identification of the active ingredients of plant and, ultimately, productivity 39. Despite all the improve­
extracts that are used in Chinese and other traditional med­ ments over the past decades towards more effective
icines. Natural products, therefore, have been an important drug discovery, the productivity challenge remains
source of chemotypes for systems-based drug discovery substantial, particularly with regard to the discovery of
approaches and are likely to have an important role in first‑in‑class drugs. In this context, phenotypic screen­
future phenotypic screening as well as in chemocentric ing could be an important contribution as it offers the
approaches. potential to provide important pharmacological tools
The distribution of first‑in‑class drugs across target to study new biology 34–37 at a faster pace than classical
families is similar to that previously published for all chemocentric systems-based approaches. We would like
drugs33, with G protein-coupled receptors (GPCRs), to emphasize that, in our view, the distinction between
kinases, proteases and ion channels being the major tar­ phenotypic screening and chemocentric drug discovery
get families in addition to the ‘other enzymes’ category, is not just a semantic one; rather, phenotypic screening
which contains a number of smaller target families and as we define it here is a new discipline. A plea for more
singletons (FIG. 5b). Probably based on the lack of suit­ phenotypic screening in drug discovery, as has been
able chemotypes before the 1980s, kinase and protease made frequently during recent years and is being imple­
inhibitors are under-represented or absent from the list mented in many groups40–44, should not be taken as a
of systems-based drugs, and were predominantly dis­ call to revert to the classical, chemocentric approach to
covered by target-based approaches. By contrast, drugs drug discovery.
that target ion channels are over-represented in this Phenotypic screening holds the promise to uncover
list, perhaps reflecting some of the difficulty and com­ new therapeutic principles and molecular pathways of cur­
plexity in discovering such inhibitors in a target-based, rently untreatable diseases45. Indeed, a number of highly
rational manner. encouraging recent examples of potential drugs derived
Interestingly, for nine (26%) of the 33 systems-based from phenotypic screening 46–51, such as bromodomain
drugs, the mechanism of action is unclear or even inhibitors50 and hepatitis C virus NS5A inhibitors51, are
unknown, which highlights that knowledge of the mech­ now in late-stage clinical trials. However, so far only a
anism of action might be helpful but is not mandatory few of the approved drugs were discovered through
to successfully develop a drug. However, elucidating the phenotypic screening, the majority of them either
mechanism of action of molecules identified through sys­ being anti-infective or anti-proliferative compounds
tems-based approaches by using forward chemical genet­ (see above).
ics, chemoproteomics or other chemical biology methods Reporter gene assays are a specific category of pheno­
could access a large untapped potential for the discovery typic screens, although they are somewhat artificial and
of novel mechanisms and therapeutic principles34–37. have limitations52,53. Many reporter gene assays have
In this way, systems- and target-based approaches are been run in pharmaceutical companies and academic
often interlinked, and what started as a systems-based institutions over the past 20 years, with a peak in the mid
drug discovery effort might uncover important tool to late 1990s, apparently with little success in terms of
compounds for further target-based approaches. The dis­ delivering drug candidates. We did not find any example
covery of the mammalian target of rapamycin (mTOR) of a first‑in‑class drug originating from such a screen,
pathway, for example, was greatly facilitated by studies but noticed several examples of target-based drugs that
with sirolimus38, and there are numerous other examples. could have been identified in such assays.
One of the fundamental challenges of phenotypic
Phenotypic screening as a new discipline screening is the selection of a few interesting compounds
In the course of almost three decades of target-based from a large list of active substances, which typically con­
drug discovery, a number of evolutionary steps have tains thousands of compounds and is often heavily dom­
been taken to improve its efficiency, and some initial inated by unselective or toxic compounds, substances
problems have been addressed. In particular, there was with unwanted mechanisms of action or false positives.
a widespread trend in the field during the late 1990s and Usually, potency is the sole criterion by which such hit lists
early 2000s to industrialize drug discovery using high- are sorted and compounds selected for further studies,
throughput methodologies in biology and chemistry. but this criterion might be inappropriate for identify­
Consequently, this approach to candidate drug discovery ing the best chemotypes. It is important to understand
was established as a linear sequence of separate steps — that we will require new methodologies and approaches
target identification, tool production and assay develop­ to increase the success rate of phenotypic screens. For
ment, hit finding and validation, hit‑to‑lead progression, example, many diseases today are still difficult to faith­
lead optimization and preclinical development — each of fully mirror in test plates and model organisms, and the
which was the focus of optimization efforts with the goal use of stem cell technology or whole-organism screens
of increasing the throughput and/or efficiency of each might enable the establishment of more physiologi­
step. The assumption was that brute force and ever-larger cal assay systems that better reflect the actual disease.
numbers of projects and high-throughput experiments Progress here will certainly increase the likelihood of
would increase productivity. finding disease-relevant pathways, but it may not enable

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 9

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

a more effective selection of interesting compounds, We therefore consider phenotypic screening not as a
and so more innovation is required. Consequently, in neoclassical approach that reverts to a supposedly more
order for phenotypic screening to be successful, it will be successful systems-based method of the past, but instead
important to recognize it as a new discipline that needs as a logical evolution of the current target-based activi­
new technologies and methods, and for which lessons ties in drug discovery. Moreover, phenotypic screening
must be learnt to avoid frustration from unrealistic is not just dependent on the use of many tools that have
expectations and premature conclusions if the current been established for target-based approaches; it also
investments do not yield quick returns. requires further technological advancements.
Perhaps we are in a phase today similar to the one The choice should not be between phenotypic screen­
in the mid‑1980s, when systems-based chemocentric ing or target-based discovery, as both approaches can
drug discovery was largely replaced by target-based complement each other, are interconnected and should
approaches. This allowed the field to greatly expand be run in parallel54. Although target or pathway discovery
beyond the relatively limited number of scaffolds that for chemotypes identified from phenotypic screens is
had been studied for decades and to gain access to many not necessarily required for further drug development,
more pharmacologically active compound classes, pro­ it could be an advantage in order to discover additional
viding a boost to innovation. Now, with an increased pathway nodes for target-based therapeutic intervention
chemical space, the time might be right to further or to enable the discovery of follow-on drugs. Therefore,
broaden the target space and open up new avenues. the goal will be to screen phenotypically in an efficient
This could well be achieved by investing in phenotypic and effective manner and to combine phenotypic screen­
screening using the compound libraries that have been ing sensibly and productively with target-based drug
established in the context of target-based approaches. discovery.

1. Drews, J. & Ryser, S. The role of innovation in drug 16. Zheng, W., Thorne, N. & McKew, J. C. Phenotypic 33. Hopkins, A. L. & Groom, C. R. The druggable
development. Nature Biotech. 15, 1318–1319 (1997). screens as a renewed approach for drug discovery. genome. Nature Rev. Drug Discov. 1, 727–730
2. Garnier, J. P. Rebuilding the R&D engine in big Drug Discov. Today 18, 1067–1073 (2013). (2002).
pharma. Harv. Bus. Rev. 86, 68–70 (2008). 17. Swinney, D. C. Phenotypic versus target-based drug 34. Lokey, R. S. Forward chemical genetics: progress
3. Douglas, F. L., Narayanan, V. K., Mitchell, L. & discovery for first‑in‑class medicines. Clin. Pharmacol. and obstacles on the path to a new pharmacopoeia.
Litan, R. E. The case for entrepreneurship in R&D in Ther. 93, 299–301 (2013). Curr. Opin. Chem. Biol. 7, 91–96 (2003).
the pharmaceutical industry. Nature Rev. Drug Discov. 18. Schrör, K. Acetylsalicylic Acid (Wiley, 2009). 35. Hall, S. E. Chemoproteomics-driven drug discovery:
9, 683–689 (2010). 19. Sneader, W. Drug Discovery: A History (Wiley, 2005). addressing high attrition rates. Drug Discov. Today 11,
4. David, E., Tramontin, T. & Zemmel, R. Pharmaceutical 20. Macarron, R. et al. Impact of high-throughput 495–502 (2006).
R&D: the road to positive returns. Nature Rev. screening in biomedical research. Nature Rev. 36. Pruss, R. M. Phenotypic screening strategies for
Drug Discov. 8, 609–610 (2009). Drug Discov. 10, 188–195 (2011). neurodegenerative diseases: a pathway to discover
5. Paul, S. M. et al. How to improve R&D productivity: 21. Beck, A., Wurch, T., Bailly, C. & Corvaia, N. novel drug candidates and potential disease targets
the pharmaceutical industry’s grand challenge. Strategies and challenges for the next generation of or mechanisms. CNS Neurol. Disord. Drug Targets 9,
Nature Rev. Drug Discov. 9, 203–214 (2010). therapeutic antibodies. Nature Rev. Immunol. 10, 693–700 (2010).
6. Kola, I. & Landis, J. Can the pharmaceutical industry 345–352 (2010). 37. St Onge, R., Schlecht, U., Scharfe, C. & Evangelista, M.
reduce attrition rates? Nature Rev. Drug Discov. 3, 22. Algire, G. H. & Chalkley, H. W. Vascular reactions of Forward chemical genetics in yeast for discovery of
711–715 (2004). normal and malignant tissues in vivo.1. Vascular chemical probes targeting metabolism. Molecules 17,
7. Ruffolo, R. R. Why has R&D productivity declined reactions of mice to wounds and to normal and 13098–13115 (2012).
in the pharmaceutical industry? Expert Opin. neoplastic transplants. J. Natl Cancer Inst. 6, 73–85 38. Heitman, J., Movva, N. R. & Hall, M. N.
Drug Discov. 1, 99–102 (2006). (1945). Targets for cell cycle arrest by the immuno­
8. Scannell, J. W., Blanckley, A., Boldon, H. & 23. Greenblatt, M. & Shubk, P. Tumor angiogenesis — suppressant rapamycin in yeast. Science 253,
Warrington, B. Diagnosing the decline in transfilter diffusion studies in hamster by transparent 905–909 (1991).
pharmaceutical R&D efficiency. Nature Rev. chamber technique. J. Natl Cancer Inst. 41, 111–124 39. Sams-Dodd, F. Is poor research the cause of the
Drug Discov. 11, 191–200 (2012). (1968). declining productivity of the pharmaceutical industry?
This article discusses four factors as the primary 24. Folkman, J. Tumor angiogenesis — therapeutic An industry in need of a paradigm shift. Drug Discov.
causes of the decline in pharmaceutical R&D implications. N. Engl. J. Med. 285, 1182–1186 Today 18, 211–217 (2013).
efficiency. (1971). This paper discusses potential reasons for the
9. Munos, B. Lessons from 60 years of pharmaceutical 25. Senger, D. R. et al. Tumor cells secrete a vascular high failure rate of target-based drug discovery
innovation. Nature Rev. Drug Discov. 8, 959–968 permeability factor that promotes accumulation of and suggests changes to improve productivity.
(2009). ascites fluid. Science 219, 983–985 (1983). 40. Lee, J. A. & Berg, E. L. Neoclassic drug discovery:
10. Horrobin, D. F. Realism in drug discovery — could 26. Leung, D. W., Cachianes, G., Kuang, W. J., the case for lead generation using phenotypic
Cassandra be right? Nature Biotech. 19, 1099–1100 Goeddel, D. V. & Ferrara, N. Vascular endothelial and functional approaches. J. Biomol. Screen. 18,
(2001). growth factor is a secreted angiogenic mitogen. 1143–1155 (2013).
11. Pammolli, F., Magazzini, L. & Riccaboni, M. Science 246, 1306–1309 (1989). 41. Carragher, N. O., Brunton, V. G. & Frame, M. C.
The productivity crisis in pharmaceutical R&D. 27. Presta, L. G. et al. Humanization of an anti-vascular Combining imaging and pathway profiling: an
Nature Rev. Drug Discov. 10, 428–438 (2011). endothelial growth factor monoclonal antibody for alternative approach to cancer drug discovery.
12. Swinney, D. C. & Anthony, J. How were new medicines the therapy of solid tumors and other disorders. Drug Discov. Today 17, 203–214 (2012).
discovered? Nature Rev. Drug Discov. 10, 507–519 Cancer Res. 57, 4593–4599 (1997). 42. Lee, J. A., Uhlik, M. T., Moxham, C. M., Tomandl, D.
(2011). 28. Buchdunger, E. et al. Inhibition of the Abl & Sall, D. J. Modern phenotypic drug discovery is a
This article analyses the origins of first‑in‑class drugs protein-tyrosine kinase in vitro and in vivo by a viable, neoclassic pharma strategy. J. Med. Chem. 55,
(FDA approvals from 1999 to 2008) and postulates 2‑phenylaminopyrimidine derivative. Cancer Res. 4527–4538 (2012).
that target-based drug discovery contributes to low 56, 100–104 (1996). 43. Chatterjee, A. K. & Yeung, B. K. Back to the future:
productivity in pharmaceutical R&D. 29. Nowell, P. C. & Hungerford, D. A minute chromosome lessons learned in modern target-based and whole-cell
13. Butcher, E. C. Can cell systems biology rescue drug in human chronic granulocytic leukemia. Science 132, lead optimization of antimalarials. Curr. Top. Med. Chem.
discovery? Nature Rev. Drug Discov. 4, 461–467 1497 (1960). 12, 473–483 (2012).
(2005). 30. Rowley, J. D. New consistent chromosomal 44. Payne, D. J., Gwynn, M. N., Holmes, D. J. &
14. Hellerstein, M. K. A critique of the molecular target- abnormality in chronic myelogenous leukemia Pompliano, D. L. Drugs for bad bugs: confronting the
based drug discovery paradigm based on principles identified by quinacrine fluorescence and Giemsa challenges of antibacterial discovery. Nature Rev.
of metabolic control: advantages of pathway-based staining. Nature 243, 290–293 (1973). Drug Discov. 6, 29–40 (2007).
discovery. Metab. Eng. 10, 1–9 (2008). 31. Shtivelman, E., Lifshitz, B., Gale, R. P. & Canaani, E. 45. Fishman, M. C. & Porter, J. A. Pharmaceuticals: a new
15. Kell, D. B. Finding novel pharmaceuticals in the Fused transcript of abl and bcr genes in chronic grammar for drug discovery. Nature 437, 491–493
systems biology era using multiple effective drug myelogenous leukemia. Nature 315, 550–554 (1985). (2005).
targets, phenotypic screening and knowledge of 32. Burnett, D. A. et al. 2‑Azetidinones as inhibitors of 46. Welch, E. M. et al. PTC124 targets genetic disorders
transporters: where drug discovery went wrong and cholesterol absorption. J. Med. Chem. 37, 1733–1736 caused by nonsense mutations. Nature 447, 87–91
how to fix it. FEBS J. 280, 5957–5980 (2013). (1994). (2007).

10 | ADVANCE ONLINE PUBLICATION www.nature.com/reviews/drugdisc

© 2014 Macmillan Publishers Limited. All rights reserved

A N A LY S I S

47. Rottmann, M. et al. Spiroindolones, a potent 55. Ogbourne, S. M. et al. Antitumor activity of 3‑ingenyl 62. Katz, D. H., Marcelletti, J. F., Khalil, M. H., Pope, L. E.
compound class for the treatment of malaria. angelate: plasma membrane and mitochondrial & Katz, L. R. Antiviral activity of 1‑docosanol, an
Science 329, 1175–1180 (2010). disruption and necrotic cell death. Cancer Res. 64, inhibitor of lipid-enveloped viruses including herpes
48. Butchbach, M. E. et al. Effects of 2,4‑diaminoquina­ 2833–2839 (2004). simplex. Proc. Natl Acad. Sci. USA 88, 10825–10829
zoline derivatives on SMN expression and phenotype 56. Lock, E. A. et al. From toxicological problem to (1991).
in a mouse model for spinal muscular atrophy. therapeutic use: the discovery of the mode of 63. Szelenyi, I. Flupirtine, a re‑discovered drug, revisited.
Hum. Mol. Genet. 19, 454–467 (2010). action of 2-(2‑nitro‑4‑trifluoromethylbenzoyl)- Inflamm. Res. 62, 251–258 (2013).
49. Chen, B. et al. Small molecule-mediated disruption of 1,3‑cyclohexanedione (NTBC), its toxicology and
Wnt-dependent signaling in tissue regeneration and development as a drug. J. Inherit. Metab. Dis. 21, Acknowledgements
cancer. Nature Chem. Biol. 5, 100–107 (2009). 498–506 (1998). The authors are grateful to I. Jones for expert help in statistical
50. Filippakopoulos, P. et al. Selective inhibition of BET 57. Coe, J. W. et al. Varenicline: an α4β2 nicotinic analysis, and thank their colleagues at the Novartis Institutes
bromodomains. Nature 468, 1067–1073 (2010). receptor partial agonist for smoking cessation. for Biomedical Research for the stimulating discussions related
51. Gao, M. et al. Chemical genetics strategy identifies J. Med. Chem. 48, 3474–3477 (2005). to this work. The authors also thank M. C. Fishman for making
an HCV NS5A inhibitor with a potent clinical effect. 58. Kessel, D., Hall, T. C. & Wodinsky, I. Transport and valuable suggestions on the content and scope of the analysis,
Nature 465, 96–100 (2010). phosphorylation as factors in the antitumor action and U. Eder for important insights into general aspects of
This paper describes the discovery of daclatasvir of cytosine arabinoside. Science 156, 1240–1241 pharmaceutical research.
(BMS‑790052) by phenotypic screening and (1967).
Competing interests statement
further chemical optimization in the absence 59. Gudas, L. J., Ullman, B., Cohen, A. & Martin, D. W. Jr.
The authors declare competing interests: see Web version for
of target knowledge. Deoxyguanosine toxicity in a mouse T lymphoma:
details.
52. Naylor, L. H. Reporter gene technology: the future looks relationship to purine nucleoside phosphorylase-
bright. Biochem. Pharmacol. 58, 749–757 (1999). associated immune dysfunction. Cell 14, 531–538
53. Chiba, T., Tsuchiya, T., Mori, R. & Shimokawa, I. Protein (1978). DATABASES
reporter bioassay systems for the phenotypic screening 60. Cohen, A., Lee, J. W. & Gelfand, E. W. Drugs@FDA: http://www.accessdata.fda.gov/Scripts/cder/
of candidate drugs: a mouse platform for anti-aging Selective toxicity of deoxyguanosine and arabinosyl drugsatfda/index.cfm
drug screening. Sensors 12, 1648–1656 (2012). guanine for T‑leukemic cells. Blood 61, 660–666
54. Williams, M. Systems and integrative biology as (1983). SUPPLEMENTARY INFORMATION
alternative guises for pharmacology: prime time 61. Lambe, C. U. et al. 2‑amino‑6‑methoxypurine See online article: S1 (table) | S2 (box)
for an iPharm concept? Biochem. Pharmacol. 70, arabinoside: an agent for T‑cell malignancies. ALL LINKS ARE ACTIVE IN THE ONLINE PDF
1707–1716 (2005). Cancer Res. 55, 3352–3356 (1995).

NATURE REVIEWS | DRUG DISCOVERY ADVANCE ONLINE PUBLICATION | 11

© 2014 Macmillan Publishers Limited. All rights reserved