Recognition of general patterns using neural networks

The Hopfield model of neural network stores memory in its symmetric synaptic connections and can only learn to recognize sets of nearly orthogonal patterns. A new algorithm is put forth to permit the recognition of general (non-orthogonal) patterns. The algorithm specifies the construction of the new network's memory matrix T _ij, which is, in general, asymmetrical and contains the Hopfield neural network (Hopfield 1982) as a special case. We find further that in addition to this new algorithm for general pattern recognition, there exists in fact a large class of T _ij memory matrices which permit the recognition of non-orthogonal patterns. The general form of this class of T _ij memory matrix is presented, and the projection matrix neural network (Personnaz et al. 1985) is found as a special case of this general form. This general form of memory matrix extends the library of memory matrices which allow a neural network to recognize non-orthogonal patterns. A neural network which followed this general form of memory matrix was modeled on a computer and successfully recognized a set of non-orthogonal patterns. The new network also showed a tolerance for altered and incomplete data. Through this new method, general patterns may be taught to the neural network.     

Species-specific TNF induction of thymocyte proliferation

Summary Recombinant murine (rMu) tumor necrosis factor (TNF), in a standard comitogenic assay with phytohemagglutinin, induced murine thymocyte proliferation, while up to 10,000-fold higher concentrations of recombinant human TNF did not. The induction of thymocyte proliferation was dependent upon TNF concentration in a biphasic manner. Thus, 100 to 1000 units/ml TNF were near optimal while concentrations 1,000 units/ml caused apparent down regulation. The effect was abrogated by neutralizing antibody to rMu-TNF but not by neutralizing antibody to rMu-interleukin 1 or . The rMu-TNF did not induce proliferation of the mature murine T-helper cell line, D10.G4.1, in the presence of mitogen. Taken together the results indicate that TNF, in a strictly species-specific manner, can regulate thymocyte proliferation independently of interleukin 1.Supported in part by Asahi Chemical Industry Co., Inc. and by USPHS Grants CA-24538, CA-15142 and CA-09072 awarded by the National Cancer Institute, Department of Health and Human Services     

Analysis for linkage between F13A and three chromosome 6 marker loci: evidence for 6pter: F13A:HLA:GL01:cen gene order

Summary The results of the present study provide independent support for F13A:HLA linkage and refine the F13A: HLA and F13A: GLO1 linkage relationships. Analysis of the corresponding recombination fractions for the total paternal F13A:HLA and F13A:GLO1 peak lod scores() indicates a locus order of 6pter: F13A:HLA:GLO1:cen. Lod scores between F13A and PLG, a locus recently assigned to chromosome 6, exclude close linkage between these loci.     

An amino-terminal fragment of cholecystokinin-58 is present in the gut: evidence for a similar processing site of procholecystokinin in canine gut and brain

Radioimmunoassays using antibodies specific for the carboxyl terminus of cholecystokinin (CCK) and the midportion of CCK-58 (raised against synthetic canine CCK-33-(1-27] revealed the existence of a CCK fragment in canine gut and brain extracts which lacks the biologically active carboxyl terminal immunoreactivity. This material eluted on Sephadex G-50 gel permeation chromatography in the region of CCK-58, on high-pressure liquid chromatography (HPLC) after CCK-39 and before CCK-58, and on cation-exchange FPLC it eluted after CCK-58. The immunoreactive pattern, the ratio of absorbance at 280-220 nm and the chromatographic elution positions suggest that this large CCK-like molecule represents an amino-terminal fragment of CCK-58. This fragment is present in canine gut and brain. Therefore, a similar processing site of procholecystokinin is suggested in both tissues.     

Estimation of oxygen penetration depth in immobilized cells

Summary A simple method is proposed for calculating oxygen pentration depth in immobilized cells by assuming zero order kinetics in the presence of several external oxygen transport resistances. Calculations indicate that typical penetration depths of oxygen for immobilized microbial cells are in the range of 50–200 and those for immobilized or encapsulated animal and plant tissue culture are about 500–1000 . Based on calculations, oxygen transport in microencapsulation and microcarriers for tissue cultures are not transport-limited, but a slight limitation is expected for those in a hollow fiber reactor.Nomenclature a_s specific area of a support (cm) - Bi Biot number - dimensionless - C_b oxygen concentration in the bulk liquid (mM) - C _b C _b ^* -C_cr (mM) - C _b ^* bulk oxygen concentration in equilibrium with air (mM) - C_cr critical oxygen concentration (mM) - C_s oxygen concentration in the solid phase (mM) - d_p diameter or thickness of a support (cm) - D_eff effective diffusivity of oxygen in the solid phase (cm²/s) - k_m membrane permeability of oxygen (cm/s) - k _m ^* D_eff/_m - k_La_L liquid phase mass transfer rate coefficient (1/s) - k_sa_s solid phase mass transfer rate coefficient (1/s) - (OUR)_v volumetric oxygen uptake rate (mmol O₂/l) - p geometry parameter, p=0 for slab, p=1 for cylinder, p=2 for sphere - P_d oxygen penetration depth (cm) - P _d oxygen penetration depth in the absence of external diffusion limitation (cm) - Q volumetric oxygen uptake rate, (mmol O₂/l·h) - specific oxygen uptake rate (mmol O₂gm biomass (dry)·h) - r length coordinate (cm) - r_c oxygen penetration depth for sphere (cm) - r _c r_c in the absence of external diffusion limitation (cm) - r _c ^* oxygen penetration depth for cylinder (cm) - r _c ^* r _c ^* in the absence of external diffusion limitation (cm) - r_com combined mass transfer rate resistance (s) - r_d location where C_s becomes zero or C_cr (cm) - r_i radius of cylinder or sphere, half thickness of slab (cm) - U_sg superficial gas velocity (cm/s) - X cell concentration (g/l) Greek letters Thiele modulus, dimensionless - _L, _s liquid and solid phase volume fraction, respectively, dimensionless - effectiveness factor On sabbatical leave from KAIST, Seoul, Korea     

Isolation and characterization of rabbit gastrin

The heptadecapeptide form the rabbit gastrin was extracted from 16 rabbit antra and purified by a combination of DEAE Sephadex, C-18 SEP PAK cartridges, fast performance liquid chromatography (FPLC) and reverse phase high pressure liquid chromatography (HPLC) steps. After the HPLC purification, a sharp, single peak of gastrin-like immunoreactivity was detected that had the same absorption to immunoreactivity ratio as human gastrin. An amino terminal pyrrolidone carboxylic acid blocking group was removed by incubation with pyrrolidone carboxylic peptidase. The amino acid analysis, microsequence analysis and mass spectrometry all confirmed the structure of rabbit gastrin being pQGPWLQEEEEAYGWMDFamide. This sequence is identical to human gastrin-17 except for glutamine in position 6 which replaces glutamate in human gastrin. Both sulfated and unsulfated rabbit gastrin-17 were characterized by mass spectrometry.     

Chemical characterization of rat cholecystokinin-58

Cholecystokinin-58 (CCK-58) was purified from rat intestines using an extraction method that yields large amounts of this peptide. Greater than 30% of total CCK immunoreactivity eluted before CCK-39 upon gel permeation chromatography (Sephadex G-50) if extracts were loaded onto Sep Pak cartridges before freezing. If the extracts were frozen and stored at −70°C for six weeks, only 20% of the material eluted in this region and total immunoreactivity was reduced by 50%, suggesting that proteases were active under these storage conditions. This early eluting peak was purified by reverse phase and ion-exchange HPLC to a single absorbance peak. Microsequence analysis of this peak detected AVLRPDSEP which is the amino terminus of rat CCK-58 predicted from the rat preprocholecystokinin cDNA. Because degradation of CCK-58 occurred in these extracts, it is possible that CCK-58 is the predominant molecule form in the rat small intestine.     

[3H]MK-801 Labels a Site on the N-Methyl-D-Aspartate Receptor Channel Complex in Rat Brain Membranes

The potent noncompetitive N-methyl-D-aspartate (NMDA) receptor antagonist [3H]MK-801 bound with nanomolar affinity to rat brain membranes in a reversible, saturable, and stereospecific manner. The affinity of [3H]MK-801 was considerably higher in 5 mM Tris-HCl (pH 7.4) than in previous studies using Krebs-Henseleit buffer. [3H]MK-801 labels a homogeneous population of sites in rat cerebral cortical membranes with KD of 6.3 nM and Bmax of 2.37 pmol/mg of protein. This binding was unevenly distributed among brain regions, with hippocampus greater than cortex greater than olfactory bulb = striatum greater than medulla-pons, and the cerebellum failing to show significant binding. Detailed pharmacological characterization indicated [3H]MK-801 binding to a site which was competitively and potently inhibited by known noncompetitive NMDA receptor antagonists, such as phencyclidine, thienylcyclohexylpiperidine (TCP), ketamine, N-allylnormetazocine (SKF 10,047), cyclazocine, and etoxadrol, a specificity similar to sites labelled by [3H]TCP. These sites were distinct from the high-affinity sites labelled by the sigma receptor ligand (+)-[3H]SKF 10,047. [3H]MK-801 binding was allosterically modulated by the endogenous NMDA receptor antagonist Mg2+ and by other active divalent cations. These data suggest that [3H]MK-801 labels a high-affinity site on the NMDA receptor channel complex, distinct from the NMDA recognition site, which is responsible for the blocking action of MK-801 and other noncompetitive NMDA receptor antagonists.