Transforming neonatal care with artificial intelligence: challenges, ethical consideration, and opportunities

Artificial intelligence (AI) offers tremendous potential to transform neonatology through improved diagnostics, personalized treatments, and earlier prevention of complications. However, there are many challenges to address before AI is ready for clinical practice. This review defines key AI concepts and discusses ethical considerations and implicit biases associated with AI. Next we will review literature examples of AI already being explored in neonatology research and we will suggest future potentials for AI work. Examples discussed in this article include predicting outcomes such as sepsis, optimizing oxygen therapy, and image analysis to detect brain injury and retinopathy of prematurity. Realizing AI's potential necessitates collaboration between diverse stakeholders across the entire process of incorporating AI tools in the NICU to address testability, usability, bias, and transparency. With multi-center and multi-disciplinary collaboration, AI holds tremendous potential to transform the future of neonatology.

Clinical and economic impacts of a modified-observational screening approach to well-appearing infants born to mothers with chorioamnionitis

OBJECTIVE

Term infants born to mothers with chorioamnionitis are at risk for early-onset sepsis (EOS). We aimed to measure the impact of changing from a categorical to a modified-observational EOS screening approach on NICU admission, antibiotic utilization, and hospitalization costs.

STUDY DESIGN

Single-center retrospective pre-post cohort study of full-term infants born to mothers with chorioamnionitis. Primary outcomes included NICU admission, antibiotic utilization, and hospitalization costs. Outcomes were adjusted for demographic variables. Budget-impact analysis was performed using bootstrapping with replication.

RESULTS

380 term infants were included (197 categorical; 183 modified-observational). There was a significant decrease in NICU admission and antibiotic utilization (p < 0.05) in the modified-observational cohort but no significant difference in per-patient total hospitalization costs. Budget-impact analysis suggested a high probability of cost savings.

CONCLUSION

A modified-observational approach to evaluating term infants of mothers with chorioamnionitis can reduce NICU admission and unnecessary antibiotic therapy, and may lead to cost-savings.

A group B Streptococcus indexed transposon mutant library to accelerate genetic research on an important perinatal pathogen

IMPORTANCE

Group B Streptococcus (GBS) is a significant global cause of serious infections, most of which affect pregnant women, newborns, and infants. Studying GBS genetic mutant strains is a valuable approach for learning more about how these infections are caused and is a key step toward developing more effective preventative and treatment strategies. In this resource report, we describe a newly created library of defined GBS genetic mutants, containing over 1,900 genetic variants, each with a unique disruption to its chromosome. An indexed library of this scale is unprecedented in the GBS field; it includes strains with mutations in hundreds of genes whose potential functions in human disease remain unknown. We have made this resource freely available to the broader research community through deposition in a publicly funded bacterial maintenance and distribution repository.

Group B Streptococcus Cas9 variants provide insight into programmable gene repression and CRISPR-Cas transcriptional effects

Group B Streptococcus (GBS; S. agalactiae) causes chorioamnionitis, neonatal sepsis, and can also cause disease in healthy or immunocompromised adults. GBS possesses a type II-A CRISPR-Cas9 system, which defends against foreign DNA within the bacterial cell. Several recent publications have shown that GBS Cas9 influences genome-wide transcription through a mechanism uncoupled from its function as a specific, RNA-programmable endonuclease. We examine GBS Cas9 effects on genome-wide transcription through generation of several isogenic variants with specific functional defects. We compare whole-genome RNA-seq from Δcas9 GBS with a full-length Cas9 gene deletion; dcas9 defective in its ability to cleave DNA but still able to bind to frequently occurring protospacer adjacent motifs; and scas9 that retains its catalytic domains but is unable to bind protospacer adjacent motifs. Comparing scas9 GBS to the other variants, we identify nonspecific protospacer adjacent motif binding as a driver of genome-wide, Cas9 transcriptional effects in GBS. We also show that Cas9 transcriptional effects from nonspecific scanning tend to influence genes involved in bacterial defense and nucleotide or carbohydrate transport and metabolism. While genome-wide transcription effects are detectable by analysis of next-generation sequencing, they do not result in virulence changes in a mouse model of sepsis. We also demonstrate that catalytically inactive dCas9 expressed from the GBS chromosome can be used with a straightforward, plasmid-based, single guide RNA expression system to suppress transcription of specific GBS genes without potentially confounding off-target effects. We anticipate that this system will be useful for study of nonessential and essential gene roles in GBS physiology and pathogenesis.

Group B Streptococcus Cas9 variants provide insight into programmable gene repression and CRISPR-Cas transcriptional effects

Group B Streptococcus (GBS; S. agalactiae ) causes chorioamnionitis, neonatal sepsis, and can also cause disease in healthy or immunocompromised adults. GBS possesses a type II-A CRISPR-Cas9 system, which defends against foreign DNA within the bacterial cell. Several recent publications have shown that GBS Cas9 influences genome-wide transcription through a mechanism uncoupled from its function as a specific, RNA-programmable endonuclease. We examine GBS Cas9 effects on genome-wide transcription through generation of several isogenic variants with specific functional defects. We compare whole-genome RNA-seq from Δ cas9 GBS with a full-length Cas9 gene deletion; dcas9 defective in its ability to cleave DNA but still able to bind to frequently occurring protospacer adjacent motifs; and scas9 that retains its catalytic domains but is unable to bind protospacer adjacent motifs. Comparing scas9 GBS to the other variants, we identify nonspecific protospacer adjacent motif binding as a driver of genome-wide, Cas9 transcriptional effects in GBS. We also show that Cas9 transcriptional effects from nonspecific scanning tend to influence genes involved in bacterial defense and nucleotide or carbohydrate transport and metabolism. While genome-wide transcription effects are detectable by analysis of next-generation sequencing, they do not result in virulence changes in a mouse model of sepsis. We also demonstrate that catalytically inactive dCas9 expressed from the GBS chromosome can be used with a straightforward, plasmid-based, single guide RNA expression system to suppress transcription of specific GBS genes without potentially confounding off-target effects. We anticipate that this system will be useful for study of nonessential and essential gene roles in GBS physiology and pathogenesis.

Ventilator-Associated Pneumonia in the Neonatal Intensive Care Unit

See Bonus NeoBriefs videos and downloadable teaching slides Intubated infants in the NICU are at risk of developing ventilator-associated pneumonia (VAP), a common type of health care-associated infection. The Centers for Disease Control and Prevention developed guidelines for diagnosing VAP in patients younger than 1 year, which include worsening gas exchange, radiographic findings, and at least 3 defined clinical signs of pneumonia. VAP in infants is treated with empiric antibiotics selected based on local resistance patterns and individualized patient data. Many NICUs have implemented prevention bundles in an effort to decrease VAP by ensuring the cleanest environment for intubated neonates (hand hygiene, sterile handling of equipment), positioning of infants to prevent gastric reflux, and constantly reevaluating for extubation readiness. Although these prevention bundle elements are intuitive and generally low risk, none are based on strong research support. This article reviews the epidemiology, pathogenesis, diagnosis, treatment, and prevention of VAP in NICU patients, focusing on recent evidence, highlighting areas of emerging research, and identifying persistent knowledge gaps.

Interpretable prediction of necrotizing enterocolitis from machine learning analysis of premature infant stool microbiota

Background

Necrotizing enterocolitis (NEC) is a common, potentially catastrophic intestinal disease among very low birthweight premature infants. Affecting up to 15% of neonates born weighing less than 1500 g, NEC causes sudden-onset, progressive intestinal inflammation and necrosis, which can lead to significant bowel loss, multi-organ injury, or death. No unifying cause of NEC has been identified, nor is there any reliable biomarker that indicates an individual patient’s risk of the disease. Without a way to predict NEC in advance, the current medical strategy involves close clinical monitoring in an effort to treat babies with NEC as quickly as possible before irrecoverable intestinal damage occurs. In this report, we describe a novel machine learning application for generating dynamic, individualized NEC risk scores based on intestinal microbiota data, which can be determined from sequencing bacterial DNA from otherwise discarded infant stool. A central insight that differentiates our work from past efforts was the recognition that disease prediction from stool microbiota represents a specific subtype of machine learning problem known as multiple instance learning (MIL).

Results

We used a neural network-based MIL architecture, which we tested on independent datasets from two cohorts encompassing 3595 stool samples from 261 at-risk infants. Our report also introduces a new concept called the “growing bag” analysis, which applies MIL over time, allowing incorporation of past data into each new risk calculation. This approach allowed early, accurate NEC prediction, with a mean sensitivity of 86% and specificity of 90%. True-positive NEC predictions occurred an average of 8 days before disease onset. We also demonstrate that an attention-gated mechanism incorporated into our MIL algorithm permits interpretation of NEC risk, identifying several bacterial taxa that past work has associated with NEC, and potentially pointing the way toward new hypotheses about NEC pathogenesis. Our system is flexible, accepting microbiota data generated from targeted 16S or “shotgun” whole-genome DNA sequencing. It performs well in the setting of common, potentially confounding preterm neonatal clinical events such as perinatal cardiopulmonary depression, antibiotic administration, feeding disruptions, or transitions between breast feeding and formula.

Conclusions

We have developed and validated a robust MIL-based system for NEC prediction from harmlessly collected premature infant stool. While this system was developed for NEC prediction, our MIL approach may also be applicable to other diseases characterized by changes in the human microbiota.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-022-04618-w.

Group B Streptococcus capsular serotype alters vaginal colonization fitness

BACKGROUND

Group B Streptococcus (GBS) remains a leading cause of infant morbidity and mortality. A candidate vaccine targets six GBS serotypes, offering a potential alternative to intrapartum antibiotic prophylaxis to reduce disease burden. However, our understanding of the contributions of specific capsule types to GBS colonization and disease remains limited.

METHODS

Using allelic exchange, we generated isogenic GBS strains differing only in the serotype-determining region in two genetic backgrounds, including the hypervirulent clonal complex (CC) 17. Using a murine model of vaginal co-colonization, we evaluated the roles of the presence of capsule and of expression of specific capsular types in GBS vaginal colonization fitness independent of other genetic factors.

RESULTS

Encapsulated wild-type strains COH1 (CC17, serotype III) and A909 (non-CC17, serotype Ia), outcompeted isogenic acapsular mutants in murine vaginal co-colonization. COH1 wild-type outcompeted A909. Notably, expression of type Ia capsule conferred an advantage over type III capsule in both genetic backgrounds.

CONCLUSIONS

Specific capsule types may provide an advantage in GBS vaginal colonization in vivo. However, success of certain GBS lineages, including CC17, likely involves both capsule and non-capsule genetic elements. Capsule switching in GBS, a potential outcome of conjugate vaccine programs, may alter colonization fitness or pathogenesis.

Can we back off using antibiotics in the NICU?

Antibiotics are extensively and inconsistently prescribed in neonatal ICUs, and usage does not correlate with rates of culture positive sepsis. There is mounting data describing the short and long-term adverse effects associated with antibiotic overuse in neonates, including the increased burden of multi-drug resistant organisms. Currently there is considerable variation in antibiotic prescribing practice among neonatologists. Applying the practice of antibiotic stewardship in the NICU is crucial for standardizing antibiotic use and improving outcomes in this population. Several approaches have been proposed to identify neonatal sepsis, with the hope of reducing antibiotic utilization. These strategies all have their limitations, and often include laboratory testing and treatment of well-appearing, non-septic, infants. A conservative "watch and wait" algorithm is suggested as an alternative method for when to initiate antibiotics. This observational approach relies on availability of trained personnel able to examine infants at specified intervals, without delaying antibiotics, should signs of sepsis arise.

Genome-Wide fitness analysis of group B Streptococcus in human amniotic fluid reveals a transcription factor that controls multiple virulence traits

Streptococcus agalactiae (group B Streptococcus; GBS) remains a dominant cause of serious neonatal infections. One aspect of GBS that renders it particularly virulent during the perinatal period is its ability to invade the chorioamniotic membranes and persist in amniotic fluid, which is nutritionally deplete and rich in fetal immunologic factors such as antimicrobial peptides. We used next-generation sequencing of transposon-genome junctions (Tn-seq) to identify five GBS genes that promote survival in the presence of human amniotic fluid. We confirmed our Tn-seq findings using a novel CRISPR inhibition (CRISPRi) gene expression knockdown system. This analysis showed that one gene, which encodes a GntR-class transcription factor that we named MrvR, conferred a significant fitness benefit to GBS in amniotic fluid. We generated an isogenic targeted deletion of the mrvR gene, which had a growth defect in amniotic fluid relative to the wild type parent strain. The mrvR deletion strain also showed a significant biofilm defect in vitro. Subsequent in vivo studies showed that while the mutant was able to cause persistent murine vaginal colonization, pregnant mice colonized with the mrvR deletion strain did not develop preterm labor despite consistent GBS invasion of the uterus and the fetoplacental units. In contrast, pregnant mice colonized with wild type GBS consistently deliver prematurely. In a sepsis model the mrvR deletion strain showed significantly decreased lethality. In order to better understand the mechanism by which this newly identified transcription factor controls GBS virulence, we performed RNA-seq on wild type and mrvR deletion GBS strains, which revealed that the transcription factor affects expression of a wide range of genes across the GBS chromosome. Nucleotide biosynthesis and salvage pathways were highly represented among the set of differentially expressed genes, suggesting that MrvR may be involved in regulating nucleotide availability.

Group B Streptococcus (GBS) is a species of Gram-positive bacteria that often colonizes the healthy adult intestinal and reproductive tracts without causing serious symptoms. During pregnancy, however, GBS can invade the pregnant uterus, where it can cause infection of the placenta, fetal membranes, and fetus—a condition known as chorioamnionitis. Chorioamnionitis is associated with serious adverse pregnancy outcomes, including stillbirth, preterm labor, and severe infection of the newborn. GBS can survive in human amniotic fluid, which is low in bacterial nutrients and contains immune molecules that limit microbial persistence, and this ability likely contributes to GBS chorioamnionitis. This study is focused on a single GBS gene that encodes a genetic regulator we called MrvR, which we show is important for GBS resistance to human amniotic fluid. Using a series of genetic techniques combined with animal models of GBS colonization and infection, we show that MrvR also plays a key role in allowing GBS to invade the bloodstream and trigger the inflammatory responses that lead to preterm labor and stillbirth. The study concludes with a survey of other GBS genes whose activity is regulated by MrvR, which seems to be an important contributor to GBS virulence.

Human Herpesvirus-6 Meningitis in a Premature Infant with Fevers: A Case and Literature Review

Human herpesvirus-6 (HHV-6) is a common virus that can cause nearly universal infection in infancy and early childhood. It typically manifests as an acute febrile illness. We describe a case of a premature infant with congenital hydrocephalus secondary to aqueductal stenosis with a ventriculoperitoneal shunt in place who developed intermittent fevers while she was admitted to the neonatal intensive care unit. She was ultimately diagnosed with acute HHV-6 meningitis. In addition to this report, we present a literature review regarding this virus’s potential modes of transmission and forms of clinical presentation in the neonatal period.

Multiple Instance Learning for Predicting Necrotizing Enterocolitis in Premature Infants Using Microbiome Data

Necrotizing enterocolitis (NEC) is a life-threatening intestinal disease that primarily affects preterm infants during their first weeks after birth. Mortality rates associated with NEC are 15-30%, and surviving infants are susceptible to multiple serious, long-term complications. The disease is sporadic and, with currently available tools, unpredictable. We are creating an early warning system that uses stool microbiome features, combined with clinical and demographic information, to identify infants at high risk of developing NEC. Our approach uses a multiple instance learning, neural network-based system that could be used to generate daily or weekly NEC predictions for premature infants. The approach was selected to effectively utilize sparse and weakly annotated datasets characteristic of stool microbiome analysis. Here we describe initial validation of our system, using clinical and microbiome data from a nested case-control study of 161 preterm infants. We show receiver-operator curve areas above 0.9, with 75% of dominant predictive samples for NEC-affected infants identified at least 24 hours prior to disease onset. Our results pave the way for development of a real-time early warning system for NEC using a limited set of basic clinical and demographic details combined with stool microbiome data.

Evaluating Newborns at Risk for Early-Onset Sepsis

Early-onset sepsis (EOS) is an important cause of neonatal morbidity. Despite extensive study, identifying at-risk newborns remains challenging, especially if they are initially well appearing. Existing official EOS recommendations suggest a conservative approach that likely results in overtreatment of a low-risk population. Recent studies indicate that more precise risk assessment and alternative management strategies could decrease the number of infants exposed to blood draws and antibiotics during evaluations for EOS. This article reviews existing guidelines and provides an overview of the Bayesian sepsis calculator and serial observation as an alternative to laboratory studies and empirical antibiotics.

Improving the Sensitivity of Real-time PCR Detection of Group B Streptococcus Using Consensus Sequence-Derived Oligonucleotides

Group B Streptococcus (GBS) is a perinatal pathogen and an emerging cause of disease in adults. Culture-independent GBS detection relies on polymerase chain reaction (PCR) of conserved genes, including sip. We demonstrate suboptimal sensitivity of the existing sip PCR strategy and validate an improved method based on consensus sequences from >100 GBS genomes.

Pneumonia

Neonatal pneumonia may occur in isolation or as one component of a larger infectious process. Bacteria, viruses, fungi, and parasites are all potential causes of neonatal pneumonia, and may be transmitted vertically from the mother or acquired from the postnatal environment. The patient's age at the time of disease onset may help narrow the differential diagnosis, as different pathogens are associated with congenital, early-onset, and late-onset pneumonia. Supportive care and rationally selected antimicrobial therapy are the mainstays of treatment for neonatal pneumonia. The challenges involved in microbiological testing of the lower airways may prevent definitive identification of a causative organism. In this case, secondary data must guide selection of empiric therapy, and the response to treatment must be closely monitored.

The essential genome of Streptococcus agalactiae

BACKGROUND

Next-generation sequencing of transposon-genome junctions from a saturated bacterial mutant library (Tn-seq) is a powerful tool that permits genome-wide determination of the contribution of genes to fitness of the organism under a wide range of experimental conditions. We report development, testing, and results from a Tn-seq system for use in Streptococcus agalactiae (group B Streptococcus; GBS), an important cause of neonatal sepsis.

METHODS

Our method uses a Himar1 mini-transposon that inserts at genomic TA dinucleotide sites, delivered to GBS on a temperature-sensitive plasmid that is subsequently cured from the bacterial population. In order to establish the GBS essential genome, we performed Tn-seq on DNA collected from three independent mutant libraries-with at least 135,000 mutants per library-at serial 24 h time points after outgrowth in rich media.

RESULTS

After statistical analysis of transposon insertion density and distribution, we identified 13.5 % of genes as essential and 1.2 % as critical, with high levels of reproducibility. Essential and critical genes are enriched for fundamental cellular housekeeping functions, such as acyl-tRNA biosynthesis, nucleotide metabolism, and glycolysis. We further validated our system by comparing fitness assignments of homologous genes in GBS and a close bacterial relative, Streptococcus pyogenes, which demonstrated 93 % concordance. Finally, we used our fitness assignments to identify signal transduction pathway components predicted to be essential or critical in GBS.

CONCLUSIONS

We believe that our baseline fitness assignments will be a valuable tool for GBS researchers and that our system has the potential to reveal key pathogenesis gene networks and potential therapeutic/preventative targets.

Diagnosis of a rare fetal haemoglobinopathy in the age of next-generation sequencing

Neonatal cyanosis resulting from a fetal methaemoglobin variant is rare. Most such variants are only described in a few published case reports. We present the case of a newborn with unexplained persistent cyanosis, ultimately determined to have a γ-chain mutation causing Hb FM-Fort Ripley. This neonatal haemoglobinopathy can be challenging to diagnose, as significant oxygen desaturation may result from barely detectable levels of the mutant haemoglobin and co-oximetry studies may show a falsely normal methaemoglobin level. Our analysis of the infant's haemoglobin included high-performance liquid chromatography, cellulose acetate electrophoresis and citrate agar electrophoresis, which showed trace amounts of a suspected variant. Ultimately, the diagnosis was made through a novel application of next-generation sequencing (NGS). NGS-based diagnostic approaches are becoming increasingly available to clinicians, and our case provides a framework and evidence for the utilisation of such testing paradigms in the diagnosis of a rare cause of neonatal cyanosis.

Healthcare-associated infections in the hospitalized neonate: a review

Healthcare-associated infections in the neonatal intensive care unit add considerably to hospital stays and costs, and contribute to numerous adverse outcomes, including death. The relatively high prevalence of healthcare-associated infections among neonates is secondary to the newborn's underdeveloped immune system, the need for frequent invasive procedures, and generally prolonged hospitalization. Central line associated bloodstream infections (CLABSI) are the most common form of healthcare-associated infection, with coagulase-negative Staphylococcus species (CONS) being the most commonly cultured microorganism. Interpretation of culture results in the setting of any suspected healthcare-associated infection can be made difficult by the possibility that a recovered organism represents a commensal contaminant, rather than an actual cause of infection. This is especially true in the case of a blood culture that grows CONS during evaluation for suspected CLABSI. This article provides an overview of the epidemiology, diagnosis, prevention, and treatment of healthcare-associated infections in the NICU.

Group B Streptococcus β-hemolysin/cytolysin breaches maternal-fetal barriers to cause preterm birth and intrauterine fetal demise in vivo

BACKGROUND

Maternal vaginal colonization with Streptococcus agalactiae (Group B Streptococcus [GBS]) is a precursor to chorioamnionitis, fetal infection, and neonatal sepsis, but the understanding of specific factors in the pathogenesis of ascending infection remains limited.

METHODS

We used a new murine model to evaluate the contribution of the pore-forming GBS β-hemolysin/cytolysin (βH/C) to vaginal colonization, ascension, and fetal infection.

RESULTS

Competition assays demonstrated a marked advantage to βH/C-expressing GBS during colonization. Intrauterine fetal demise and/or preterm birth were observed in 54% of pregnant mice colonized with wild-type (WT) GBS and 0% of those colonized with the toxin-deficient cylE knockout strain, despite efficient colonization and ascension by both strains. Robust placental inflammation, disruption of maternal-fetal barriers, and fetal infection were more frequent in animals colonized with WT bacteria. Histopathologic examination revealed bacterial tropism for fetal lung and liver.

CONCLUSIONS

Preterm birth and fetal demise are likely the direct result of toxin-induced damage and inflammation rather than differences in efficiency of ascension into the upper genital tract. These data demonstrate a distinct contribution of βH/C to GBS chorioamnionitis and subsequent fetal infection in vivo and showcase a model for this most proximal step in GBS pathogenesis.

Retrocyclin inhibits Gardnerella vaginalis biofilm formation and toxin activity

BACKGROUND

Retrocyclins are cyclic antimicrobial peptides that have been shown to be both broadly active and safe in animal models. RC-101, a synthetic retrocyclin, targets important human pathogens and is a candidate vaginal microbicide. Its activity against microbes associated with bacterial vaginosis is unknown.

METHODS

We investigated the effect of RC-101 on toxin activity, bacterial growth and biofilm formation of Gardnerella vaginalis in vitro.

RESULTS

RC-101 potently inhibits the cytolytic activity of vaginolysin, the Gardnerella vaginalis toxin, on both erythrocytes and nucleated cells. RC-101 lacks inhibitory activity against planktonic G. vaginalis but markedly decreases biofilm formation.

CONCLUSIONS

These dual properties, toxin inhibition and biofilm retardation, justify further exploration of RC-101 as a candidate agent for bacterial vaginosis prevention.

Blind cavefish and heat shock protein chaperones: a novel role for hsp90alpha in lens apoptosis

Lens apoptosis plays a central role in cavefish eye degeneration. Heat shock proteins (hsps) can regulate apoptosis; therefore, we examined the relationship between constitutive hsp70 and hsp90 expression and lens apoptosis. The model system is Astyanax mexicanus, a teleost species consisting of an eyed surface-dwelling (surface fish) form and numerous blind cave-dwelling (cavefish) forms. Optic primordia are formed in the cavefish embryo but they subsequently undergo lens apoptosis, arrest in development and degenerate. Astyanax hsp90 and hsp70 DNAs were isolated to use as probes to compare gene expression during surface fish and cavefish development. Hsp90beta, which encodes one of two hsp90 isoforms, was not expressed in the surface fish or cavefish lens, whereas hsp70 was expressed in the lens of both forms, suggesting that neither is directly involved in lens apoptosis. In contrast, hsp90alpha, the other hsp90 isoform, was expressed in the cavefish but not the surface fish lens. Hsp90alpha expression peaked shortly before the beginning of lens apoptosis in three convergent cavefish populations, suggesting a close relationship with lens apoptosis. The absence of hsp90beta in the lens allowed us to use geldanamycin and radicicol, specific inhibitors of hsp90 chaperone function, to determine whether lens cell death requires hsp90alpha expression. Both inhibitors blocked TUNEL labeling in the cavefish lens, suggesting that hsp90alpha is required for apoptosis. In contrast to their effects on the lens, these inhibitors induced TUNEL labeling in the surface epidermis, presumably due to effects on hsp90beta function, implying that the two-hsp90 isoforms may have contrasting roles in cell survival. We conclude that hsp90alpha plays a novel role in lens apoptosis and cavefish eye degeneration.

Mr. Chips 2002: new insights from an ideal-observer model of reading

The integration of visual, lexical, and oculomotor information is a critical part of reading. Mr. Chips is an ideal-observer model that combines these sources of information optimally to read simple texts in the minimum number of saccades. This model provides a computational framework for interpreting human reading saccades in both normal and low vision. The purpose of this paper is to report performance of the model for conditions emulating reading with normal vision--a visual span of nine characters, multiplicative saccade noise with a standard deviation of 30%, and texts based on three full-length children's books. Comparison of fixation locations by humans and Mr. Chips revealed: (1) that both exhibit very similar word-skipping behavior; (2) both show initial fixations near the center of words, but with a systematic difference suggestive of an asymmetry in the human visual span; and (3) differences in the pattern of refixations within words that may uncover non-optimal lexical inference by human readers. A human context effect--30% difference in mean saccade size between continuous text and random sequences of words--was very similar to the 25% effect for the model associated with a corresponding difference in the predictability of text words. Overall, our findings show that many of the complicated aspects of human reading saccades can be explained concisely by early information-processing constraints.