Classifying vertical facial deformity using supervised and unsupervised learning.

OBJECTIVES: To evaluate the potential for machine learning techniques to identify objective criteria for classifying vertical facial deformity. METHODS: 19 parameters were determined from 131 lateral skull radiographs. Classifications were induced from raw data with simple visualisation, C5.0 and Kohonen feature maps; and using a Point Distribution Model (PDM) of shape templates comprising points taken from digitised radiographs. RESULTS: The induced decision trees enable a direct comparison of clinicians' idiosyncrasies in classification. Unsupervised algorithms induce models that are potentially more objective, but their blackbox nature makes them unsuitable for clinical application. The PDM methodology gives dramatic visualisations of two modes separating horizontal and vertical facial growth. Kohonen feature maps favour one clinician and PDM the other. Clinical response suggests that while Clinician 1 places greater weight on 5 of 6 parameters, Clinician 2 relies on more parameters that capture facial shape. CONCLUSIONS: While machine learning and statistical analyses classify subjects for vertical facial height, they have limited application in their present form. The supervised learning algorithm C5.0 is effective for generating rules for individual clinicians but its inherent bias invalidates its use for objective classification of facial form for research purposes. On the other hand, promising results from unsupervised strategies (especially the PDM) suggest a potential use for objective classification and further identification and analysis of ambiguous cases. At present, such methodologies may be unsuitable for clinical application because of the invisibility of their underlying processes. Further study is required with additional patient data and a wider group of clinicians.

The LIM proteins FHL1 and FHL3 are expressed differently in skeletal muscle.

We have determined the complete mRNA sequence of FHL3 (formerly SLIM2). We have confirmed that it is a member of the family of LIM proteins that share a similar secondary protein structure, renamed as Four-and-a-Half-LIM domain (or FHL) proteins in accordance with this structure. The "half-LIM" domain is a single zinc finger domain that may represent a subfamily of LIM domains and defines this particular family of LIM proteins. The distribution of FHL mRNA expression within a variety of murine tissues is complex. Both FHL1 and FHL3 were expressed in a number of skeletal muscles while FHL2 was expressed at high levels in cardiac muscle. Localisation of FHL3 to human chromosome 1 placed this gene in the proximity of, but not overlapping with, alleles associated with muscle diseases. FHL1 and FHL3 mRNAs were reciprocally expressed in the murine C2C12 skeletal muscle cell line and this suggested that the pattern of expression was linked to key events in myogenesis.

The fourth member of the FHL family of LIM proteins is expressed exclusively in the testis.

We have determined the sequence of a novel LIM protein, termed FHL4. It was found to contain the defining secondary structural arrangement of the FHL (formerly SLIM) family of LIM proteins. This sequence was assembled using EST sequences deposited on the dbEST database and from sequencing of PCR fragments, all of which were derived from murine testis cDNA. Northern analysis of a wide range of murine tissues demonstrated that the expression of FHL4 mRNA was restricted to the testis. Using in situ hybridisation, FHL4 mRNA was found to be associated with the seminiferous epithelium and not with the interstitial tissue surrounding the seminiferous tubules. It is suggested that mFHL4 is associated with the process of spermatogenesis.

Slim defines a novel family of LIM-proteins expressed in skeletal muscle.

We have assembled the complete protein sequence of the skeletal muscle LIM-protein SLIM by aligning overlapping cDNA sequences. These cDNA sequences were identified from our own sequencing and from BLASTn searches of non-redundant cDNA databases. The predicted SLIM protein sequence included four LIM-domains and a novel single zinc finger domain located in the N-terminal region. Similar sequences to SLIM were identified and termed SLIM2 and SLIM3. The SLIM3 cDNA sequence was identified subsequently as a partial sequence of the of the LIM-protein DRAL. The number and spacing of the LIM domains was common to all three protein sequences. The mRNA for each protein was detected in human masseter muscle RNA by Northern analysis. We suggest that these proteins belong to a novel family of LIM proteins that are expressed in human skeletal muscle.

The developmental regulation of a novel muscle LIM-protein.

Using a cDNA clone derived from a human muscle library we have identified a novel and highly conserved 2.3kb homologue which is highly expressed in skeletal muscle. The partial sequence contains at least three LIM domains and shows greatest homology with the group of LIM-proteins associated with the cytoskeleton and focal adhesion plaques which include zyxin and paxillin. This homologue is maximally expressed in differentiated ovine primary muscle cultures. It is also expressed in the ovine fetus from at least 50 days of gestation and is increasingly upregulated from 120 days of gestation to 8 weeks after birth after which it declines. This period corresponds to the period of greatest muscle fibre hypertrophy and suggests a role for this homologue in either the elaboration of muscle fibre matrix anchorage or the regulation of muscle fibre hypertrophy itself.

Effect of maternal undernutrition in early gestation on the development of fetal myofibres in the guinea-pig.

A 40% restriction in maternal feed intake throughout gestation in the guinea-pig results in a 35% reduction in fetal body weight at term and a 20-25% reduction in muscle fibre number. To investigate the effect of maternal undernutrition in early gestation, four nutritional treatments were used: controls-pregnant females fed ad libitum throughout gestation; TR-fed 60% ad libitum intake throughout gestation; ER-fed 60% ad libitum for the first third of gestation (until Day 25), then ad libitum to term; LR-fed ad libitum for the first 25 days, then 60% of ad libitum to term. The LR group were complicated by a high degree of fetal resorption and early littering of viable litters. The biceps brachii and soleus muscles were removed from neonates and total muscle fibre numbers determined. In a second experiment a further 8 pregnant guinea-pigs were fed 60% ad libitum until Day 15 of gestation only, and then rehabilitated onto an ad libitum diet (VER). Of these, 5 guinea-pigs were killed at term and the remaining 3 at 45 days gestation. Fetuses and placentae were obtained from all VER animals and compared with TR and controls of a similar age. Body weights were reduced in all restricted groups at term when compared with controls (P < 0.05) by 12, 40 and 50% for VER = ER, TR and LR groups, respectively. Biceps fibre number was reduced (P < 0.05) in ER, TR and LR groups by 28, 20 and 25%, respectively, but was not affected in the VER group. Soleus fibre number was not significantly affected by any nutritional treatment. Restriction for 15 days in early gestation caused a significant 20% increase in fetal weight at 45 days' gestation compared with controls, but muscle and placental weights were not affected. Analysis of placental components at Days 45 and 65 suggested that underfeeding in early gestation and subsequent refeeding caused some placental adaptations to increase the exchange-surface area. A short period of maternal undernutrition in the first third of gestation alone (ER), therefore, resulted in a biceps brachii fibre number deficit similar to that caused by restriction throughout gestation only if the period of restriction extended as far as Day 25. Furthermore, fetal weight at term was impaired by short-term nutritional restriction in early gestation. Restriction in the last two-thirds of gestation, following an ad libitum diet in the first third, caused a reduction in biceps fibre number and had a severe effect on the maintenance of pregnancy. It is probable that undernutrition in early gestation had an indirect effect on muscle fibre number by affecting the development of the placenta. This could be avoided by nutritional rehabilitation before Day 25 of gestation, but appeared to be permanent thereafter. Undernutrition after Day 25 may have had a direct effect on the development of secondary fibres.

The effect of maternal undernutrition on the growth and development of the guinea pig placenta.

Fetal growth is known to be correlated with the size of the placenta and the exchange surface area. Reduction in the growth of the materno-fetal exchange surface areas may be a mechanism by which the effects of maternal undernutrition on fetal growth are mediated. In the compact placenta of the guinea pig the exchange surface is equivalent to the peripheral labyrinth. The effect of a 40% reduction in maternal feed intake on the growth of the peripheral labyrinth was investigated in pregnant guinea pigs between gestational days 25 and 65. Fetal and placental weights were significantly reduced in the last trimester by 32% and 38% respectively (P < 0.01). Placental efficiency in early gestation was significantly impaired in restricted animals but equivalent to ad lib. fed controls by the last trimester. The volume of the peripheral labyrinth increased as a percentage of the total placental volume with gestational age. Restricted placentae tended to be composed of a smaller volume of peripheral labyrinth tissue in early gestation. It is suggested that maternal undernutrition results in an impaired or delayed expansion of the peripheral labyrinth in early gestation causing a reduction in placental efficiency. By the last trimester the weight of the peripheral labyrinth of restricted animals was reduced by 33% (P < 0.05). The weight of the peripheral labyrinth was also significantly correlated with fetal weight is limited by the size of the peripheral labyrinth in the later stages of gestation.

Prostaglandins and the dog stomach: effects of misoprostol on the proportion of mucosa to muscle and on the proportion of different epithelial cell types.

Dogs were given an oral dose of 300 micrograms/kg/day of the prostaglandin E1 analogue, misoprostol for 11 weeks. The relative areas of the main components of the fundus and antrum per unit area of serosa were quantified. While the volume occupied by the muscle layers did not change, there was an increase in mucosal volume in both sites. The relative volume fraction of the main mucosal cell types was then determined by a point counting technique. Misoprostol very highly significantly increased the fraction of mucosal volume occupied by surface mucus cells, and also significantly decreased the volume fraction of the parietal cells.

The effects of the prostaglandin analogue, misoprostol, on cell proliferation and cell migration in the canine stomach.

Increased mucosal mass is associated with prostaglandin administration, raising the question as to whether this is the consequence of increased cell production or decreased cell loss. 3H-thymidine was injected into 30 test dogs given 300 micrograms/kg/day of the prostaglandin E1 analogue, misoprostol, orally for 11 weeks, and into 30 dogs which had been given the vehicle alone. Six test and 6 control dogs were killed 1 h after thymidine labelling, and the remainder were killed subsequently at timed intervals to determine cell migration rates and transit times. Misoprostol significantly increased the stomach weight, but nonetheless did not significantly increase the labelling index, nevertheless, the number of labelled cells per gland was significantly increased, which was a consequence of the increased gland length. When the data was expressed as a gland cell production rate significant increases were thus observed. The migration rate of cells toward the gastric lumen increased one and a half times in the misoprostol-treated group; however, as the gland length also significantly increased, there was no significant difference in the transit time. It must be concluded that the hyperplasia associated with exogenous prostaglandins is the consequence of increased cell production, not decreased cell loss, and that previous reports to the contrary are the result of failure to allow for concomitant changes in the denominator of proliferative indices.

Effects of misoprostol on cell migration and transit in the dog stomach.

Prostaglandins of the E series increase stomach mucosal mass by inducing hyperplasia, which could be the result either of increased cell production or of decreased cell loss. This report describes an investigation of the effect of the prostaglandin E1 analogue, misoprostol, on cell migration and transit. 3H-thymidine was used to label those cells synthesizing deoxyribonucleic acid in dogs that had been given an oral dose of 300 micrograms/kg per day misoprostol for 11 weeks. The animals were killed at timed intervals, and tissue from the gastric fundus was prepared for autoradiography. The distribution of labeled cells at various times after labeling was used to follow the movement of the wave of label and to calculate median cell migration rates and transit times. The migration rate of cells toward the gastric lumen was significantly increased from 1.4 +/- 0.3 to 3.6 +/- 0.6 cell positions per day in the misoprostol-treated group (p less than 0.001); however, the gland length (from the most basal mucous neck cell to the luminal surface) was also increased (from 52.1 +/- 1.1 to 74.0 +/- 1.6; p less than 0.001), thus there was no significant difference in the (transit) time taken for cells to reach the top of the gland (control, 17.5 +/- 9.8 days; test, 12.2 +/- 7.1 days).

Prostaglandins and the gastric epithelium: effects of misoprostol on gastric epithelial cell proliferation in the dog.

The effects of the methyl ester analogue of prostaglandin E1, misoprostol, on gastric epithelial cell proliferation were investigated in six dogs given 300 micrograms/kg/day of misoprostol orally for 11 weeks and in six control dogs given placebo for 11 weeks. Misoprostol treatment resulted in a 36% increase in stomach weight (p less than 0.01) and a 30% increase in the length (measured as the cell column count from the base/neck junction to the surface) of the fundic gastric glands (p less than 0.01). This mucosal hyperplasia was predominantly caused by enlargement of the foveolar region of the gland, with little change occurring in the neck or in the isthmus. The hyperplasia was the result of an increased number of mitotic (p less than 0.01) and DNA synthesising cells (p less than 0.05) in each gastric gland, which resulted in a significant increase in the gland cell production rate, from 22.5 to 42.6 cells per gland per day (p less than 0.05).