Financial returns on R&D: looking back at history, looking forward to adaptive licensing.

Investment in R&D for drugs launched in the late 1970s to early 1990s generated good returns for investors. R&D was inexpensive. Clinical trial success rates were high. Consumption was increasing. Drug prices were outstripping inflation, which raised profit margins. Tax rates were falling. However, returns on R&D have been falling since the early 1990s given rising clinical trial costs, rising trial failure rates, and lower consumption growth in developed markets. Many investors believe that average financial returns on today's R&D will be below the cost of capital, particularly if US drug price inflation moderates. Thus R&D investment by major drug companies is flat or perhaps falling in real terms. Various regulatory initiatives have tried to streamline clinical development and approval. The latest is Adaptive Licensing (AL). The near-term effect of AL on industry-level financial returns will be modest. AL will, however, be salient for decisions to invest in specific trials and may make it easier for smaller companies to fund development. AL could become more important in the long run if it helps shift industry, regulators, and payers from what has been an increasingly linear model of innovation; predicated on the ideas that basic science predicts, trials test predictions, and trial results form a complete description of a drug's attributes. History shows that many drugs become important because doctors and patients discover utility that was not initially apparent to regulators, payers, or investors. One hope for AL, therefore, is that it will bring more acceptably safe chemical diversity into real world use at lower R&D cost.

On variability in the density of corticocortical and thalamocortical connections.

Variability is an important but neglected aspect of connectional neuroanatomy. The quantitative density of the 'same' corticocortical or thalamocortical connection may vary by over two orders of magnitude between different injections of the same tracer. At present, however, the frequency distribution of connection densities is unknown. Therefore, it is unclear what kind of sampling strategies or statistical methods are appropriate for quantitative studies of connectivity. Nor is it clear if the measured variability represents differences between subjects, or if it is simply a consequence of intra-individual differences resulting from experimental technique and the exact placement of tracers relative to local spatial and laminar variation in connectivity. We used quantitative measurements of the density of a large number of corticocortical and thalamocortical connections from our own laboratories and from the literature. Variability in the density of given corticocortical and thalamocortical connections is high, with the standard deviation of density proportional to the mean. The frequency distribution is close to exponential. Therefore, analysis methods relying on the normal distribution are not appropriate. We provide an appendix that gives simple statistical guidance for samples drawn from exponentially distributed data. For a given corticocortical or thalamocortical connection density, between-individual standard deviation is 0.85 to 1.25 times the within-individual standard deviation. Therefore, much of the variability reported in conventional neuroanatomical studies (with one tracer deposited per animal) is due to within-individual factors. We also find that strong, but not weak, corticocortical connections are substantially more variable than thalamocortical connections. We propose that the near exponential distribution of connection densities is a simple consequence of 'patchy' connectivity. We anticipate that connection data will be well described by the negative binomial, a class of distribution that applies to events occurring in clumped or patchy substrates. Local patchiness may be a feature of all corticocortical connections and could explain why strong corticocortical connections are more variable than strong thalamocortical connections. This idea is supported by the columnar patterns of many corticocortical but few thalamocortical connections in the literature.

Anatomical connectivity defines the organization of clusters of cortical areas in the macaque monkey and the cat.

The number of different cortical structures in mammalian brains and the number of extrinsic fibres linking these regions are both large. As with any complex system, systematic analysis is required to draw reliable conclusions about the organization of the complex neural networks comprising these numerous elements. One aspect of organization that has long been suspected is that cortical networks are organized into 'streams' or 'systems'. Here we report computational analyses capable of showing whether clusters of strongly interconnected areas are aspects of the global organization of cortical systems in macaque and cat. We used two different approaches to analyse compilations of corticocortical connection data from the macaque and the cat. The first approach, optimal set analysis, employed an explicit definition of a neural 'system' or 'stream', which was based on differential connectivity. We defined a two-component cost function that described the cost of the global cluster arrangement of areas in terms of the areas' connectivity within and between candidate clusters. Optimal cluster arrangements of cortical areas were then selected computationally from the very many possible arrangements, using an evolutionary optimization algorithm. The second approach, non-parametric cluster analysis (NPCA), grouped cortical areas on the basis of their proximity in multidimensional scaling representations. We used non-metric multidimensional scaling to represent the cortical connectivity structures metrically in two and five dimensions. NPCA then analysed these representations to determine the nature of the clusters for a wide range of different cluster shape parameters. The results from both approaches largely agreed. They showed that macaque and cat cortices are organized into densely intra-connected clusters of areas, and identified the constituent members of the clusters. These clusters reflected functionally specialized sets of cortical areas, suggesting that structure and function are closely linked at this gross, systems level.

On imputing function to structure from the behavioural effects of brain lesions.

What is the link, if any, between the patterns of connections in the brain and the behavioural effects of localized brain lesions? We explored this question in four related ways. First, we investigated the distribution of activity decrements that followed simulated damage to elements of the thalamocortical network, using integrative mechanisms that have recently been used to successfully relate connection data to information on the spread of activation, and to account simultaneously for a variety of lesion effects. Second, we examined the consequences of the patterns of decrement seen in the simulation for each type of inference that has been employed to impute function to structure on the basis of the effects of brain lesions. Every variety of conventional inference, including double dissociation, readily misattributed function to structure. Third, we tried to derive a more reliable framework of inference for imputing function to structure, by clarifying concepts of function, and exploring a more formal framework, in which knowledge of connectivity is necessary but insufficient, based on concepts capable of mathematical specification. Fourth, we applied this framework to inferences about function relating to a simple network that reproduces intact, lesioned and paradoxically restored orientating behaviour. Lesion effects could be used to recover detailed and reliable information on which structures contributed to particular functions in this simple network. Finally, we explored how the effects of brain lesions and this formal approach could be used in conjunction with information from multiple neuroscience methodologies to develop a practical and reliable approach to inferring the functional roles of brain structures.

Neuronal population activity and functional imaging.

Human functional brain imaging detects blood flow changes that are thought to reflect the activity of neuronal populations and, thus, the responses of neurons that carry behaviourally relevant information. Since this relationship is poorly understood, we explored the link between the activity of single neurons and their neuronal population. The functional imaging results were in good agreement with levels of population activation predicted from the known effects of sensory stimulation, learning and attention on single cortical neurons. However, the nature of the relationship between population activation and single neuron firing was very surprising. Population activation was strongly influenced by those neurons firing at low rates and so was very sensitive to the baseline or 'spontaneous' firing rate. When neural representations were sparse and neurons were tuned to several stimulus dimensions, population activation was hardly influenced by the few neurons whose firing was most strongly modulated by the task or stimulus. Measures of population activation could miss changes in information processing given simultaneous changes in neurons' baseline firing, response modulation or tuning width. Factors that can modulate baseline firing, such as attention, may have a particularly large influence on population activation. The results have implications for the interpretation of functional imaging signals and for cross-calibration between different methods for measuring neuronal activity.

The smell map: is there a commonality of odour perception?

The normal perception of odour quality is poorly understood, so formulating meaningful tests of olfaction is difficult. While tests of odour discrimination and odour detection threshold have helped quantify olfactory dysfunction, there are not yet predictive relationships between sensitivity to particular odours and particular forms of olfactory dysfunction. Using 11 commonly encountered odours, 20 normosmics performed similarity ratings of odour pairs. Multidimensional scaling, a standard behavioural sciences data analysis method, was used to explore the perceptual relationships between the odours based on their pair-wise similarity ratings. Smell maps were created for each individual as was a common or archetypal map which indicated a commonality in individuals' odour perception, far greater than chance alone (P < 10(-6)). A preliminary analysis of four hyposmics suggests that they do not conform to the normosmic archetype. Future studies assessing the relationship between odours in the archetype should improve the selection of odours to be included in tests of odour discrimination.

The connectional organization of the cortico-thalamic system of the cat.

Data on connections between the areas of the cerebral cortex and nuclei of the thalamus are too complicated to analyse with naked intuition. Indeed, the complexity of connection data is one of the major challenges facing neuroanatomy. Recently, systematic methods have been developed and applied to the analysis of the connectivity in the cerebral cortex. These approaches have shed light on the gross organization of the cortical network, have made it possible to test systematically theories of cortical organization, and have guided new electrophysiological studies. This paper extends the approach to investigate the organization of the entire cortico-thalamic network. An extensive collation of connection tracing studies revealed approximately 1500 extrinsic connections between the cortical areas and thalamic nuclei of the cat cerebral hemisphere. Around 850 connections linked 53 cortical areas with each other, and around 650 connections linked the cortical areas with 42 thalamic nuclei. Non-metric multidimensional scaling, optimal set analysis and non-parametric cluster analysis were used to study global connectivity and the 'place' of individual structures within the overall scheme. Thalamic nuclei and cortical areas were in intimate connectional association. Connectivity defined four major thalamo-cortical systems. These included three broadly hierarchical sensory or sensory/motor systems (visual and auditory systems and a single system containing both somatosensory and motor structures). The highest stations of these sensory/motor systems were associated with a fourth processing system composed of prefrontal, cingulate, insular and parahippocampal cortex and associated thalamic nuclei (the 'fronto-limbic system'). The association between fronto-limbic and somato-motor systems was particularly close.

Visual motion processing in the anterior ectosylvian sulcus of the cat.

1. Neurons that are selectively sensitive to the direction of motion of elongated contours have been found in several cortical areas in many species. However, in the striate cortex of the cat and monkey, and the extrastriate posteromedial lateral suprasylvian visual area of the cat, such cells are generally component motion selective, signaling only the direction of movement orthogonal to the preferred orientation; a direction that is not necessarily the same as the motion of the entire pattern or texture of which the cell's preferred contour is part. The primate extrastriate middle temporal area is the only cortical region currently known to contain a substantial population of pattern-motion-selective cells that respond to the shared vector of motion of mixtures of contours. 2. From analyzing published data on the connectivity of the cat's cortex, we predicted that the anterior ectosylvian visual area (AEV), situated within the anterior ectosylvian sulcus, might be a higher-order motion processing area and thus likely to contain pattern-motion-selective neurons. This paper presents the results of a study on neuronal responses in AEV. 3. Ninety percent of AEV cells that responded strongly to drifting grating and/or plaid stimuli were directionally selective (directionality index > 0.5). For this group, the mean directionality index was 0.75. Moreover, 55% of these cells were unequivocally classified as pattern motion selective and only one neuron was classified as definitely component motion selective. Thus high-level pattern motion coding occurs in the cat extrastriate cortex and is not limited to the primate middle temporal area. 4. AEV contains a heterogeneous population of directionally selective cells. There was no clear relation between the degree of directional selectivity for plaids or gratings and the degree of selectivity for pattern motion or component motion. Nevertheless, 28% of the highly responsive cells were both more strongly modulated by plaids than gratings and more pattern motion selective than component motion selective. Such cells could correspond to a population of "selection units" signaling the salience of local motion information. 5. AEV lacks global retinotopic order but the preferred direction of motion of neurons (rather than axis of motion, as in the middle temporal area and the posteromedial lateral suprasylvian visual area) is mapped systematically across the cortex. Our data are compatible with AEV being a nonretinotopic, feature-mapped area in which cells representing similar parts of "motion space" are brought together on the cortical sheet.

Non-metric multidimensional scaling in the analysis of neuroanatomical connection data and the organization of the primate cortical visual system.

Neuroanatomists have established that the various gross structures of the brain are divided into a large number of different processing regions and have catalogued a large number of connections between these regions. The connectional data derived from neuroanatomical studies are complex, and reliable conclusions about the organization of brain systems cannot be drawn from considering them without some supporting analysis. Recognition of this problem has recently led to the application of a variety of techniques to the analysis of connection data. One of the techniques that we previously employed, non-metric multidimensional scaling (NMDS), appears to have revealed important aspects of the organization of the central nervous system, such as the gross organization of the whole cortical network in two species. We present here a detailed treatment of methodological aspects of the application of NMDS to connection data. We first examine in detail the particular properties of neuroanatomical connection data. Second, we consider the details of NMDS and discuss the propriety of different possible NMDS approaches. Third, we present results of the analyses of connection data from the primate visual system, and discuss their interpretation. Fourth, we study independent analyses of the organization of the visual system, and examine the relation between the results of these analyses and those from NMDS. Fifth, we investigate quantitatively the performance of a number of data transformation and conditioning procedures, as well as tied and untied NMDS analysis of untransformed low-level data, to determine how well NMDS can recover known metric parameters from artificial data. We then re-analyse real connectivity data with the most successful methods at removing the effects of sparsity, to ensure that this aspect of data structure does not obscure others. Finally, we summarize the evidence on the connectional organization of the primate visual system, and discuss the reliability of NMDS analyses of neuroanatomical connection data.

Analysis of connectivity in the cat cerebral cortex.

The mammalian cerebral cortex is innervated by a large number of corticocortical connections. The number of connections makes it difficult to understand the organization of the cortical network. Nonetheless, conclusions about the organization of cortical systems drawn from examining connectional data have often been made in a speculative and informal manner, unsupported by any analytic treatment. Recently, progress has been made toward more systematic ways of extracting organizing principles from data on the network of connections between cortical areas of the monkey. In this article, we extend these approaches to the cortical systems of the cat. We collated information from the neuroanatomical literature about the corticocortical connections of the cat. This collation incorporated 1139 reported corticocortical connections between 65 cortical areas. We have previously used an optimization technique (Scannell and Young, 1993) to analyze this database in order to represent the connectional organization of cortical systems in the cat. Here, we report the connectional database and analyze it in a number of further ways. First, we employed rules from Felleman and Van Essen (1991) to investigate hierarchical relations among the areas. Second, we compared quantitatively the results of the optimization method with the results of the hierarchical method. Third, we examined quantitatively whether simple connection rules, which may reflect the development and evolution of the cortex, can account for the experimentally identified corticocortical connections in the database. The results showed, first, that hierarchical rules, when applied to the cat visual system, define a largely consistent hierarchy. Second, in both auditory and visual systems, the ordering of areas by hierarchical analysis and by optimization analysis was statistically significantly related. Hence, independent analyzes concur broadly in their ordering of areas in the cortical hierarchies. Third, the majority of corticocortical connections, and much of the pattern of connectivity, were accounted for by a simple "nearest-neighbor-or-next-door-but-one" connection rule, which may suggest one of the mechanisms by which the development of cortical connectivity is controlled.

Analysis of connectivity: neural systems in the cerebral cortex.

The mammalian cerebral cortex is composed of many distinct areas, which are very richly interconnected. The very large number of connections between cortical areas require analysis to be undertaken before reliable conclusions about the organization of neural systems in the cortex can be drawn. We review the methodology and results of two means of analysing central nervous connectivity, hierarchical analysis and optimization analysis. We conclude that these methods are reliable methods for analysing neural connectivity data, and that their results concur. The analyses indicate that all major cortical sensory systems are organized hierarchically, some central sensory systems are divided structurally into several "streams" of processing, the cortical motor system is embedded in the cortical somatosensory system, the frontal and limbic structures are connectionally associated, and that these frontal and limbic areas are invariably associated with the least peripheral sensory processing regions, and are therefore connectionally central. Finally, we discuss the differences on this common plan between the organizations of the cat and primate that these analyses reveal.

The connectional organization of neural systems in the cat cerebral cortex.

BACKGROUND: The mammalian brain consists of the cerebral cortical sheet, which is composed of many distinct areas, the cerebellar cortex, and many non-cortical nuclei. Powerful neuroanatomical techniques have revealed a large number of connections between these structures. The large number of brain structures and the very many connections between them form a strikingly complex network. The complexity of this network has made it difficult to understand how the central nervous system is organized. Recently, however, optimization analysis of an important subset of central nervous connections that occur between the different areas of the cerebral cortex has produced understandable and quantitative representations of the organization of cortical systems of the primate brain. RESULTS: Here we briefly report the extension of this approach to the cortical systems of the cat. There were four connectional clusters of cortical areas in the cat. These clusters of areas corresponded to the visual, auditory, and somato-motor systems, and to the frontal and limbic areas, which we call the fronto-limbic complex. All the major sensory systems were hierarchically organized, and their 'higher' stations were more closely associated with the fronto-limbic complex than were their 'lower' stations. CONCLUSIONS: Features of the organization of the cat brain, together with earlier primate results, suggest that there may be a common cortical plan in mammals. We suggest that this common plan may involve relatively discrete, hierarchically organized, cortical sensory systems and a topologically central fronto-limbic complex. Specific variations on this wiring plan may relate to evolutionary history and selection for particular ecological niches.