Use of machine learning to shorten observation-based screening and diagnosis of autism.

The Autism Diagnostic Observation Schedule-Generic (ADOS) is one of the most widely used instruments for behavioral evaluation of autism spectrum disorders. It is composed of four modules, each tailored for a specific group of individuals based on their language and developmental level. On average, a module takes between 30 and 60 min to deliver. We used a series of machine-learning algorithms to study the complete set of scores from Module 1 of the ADOS available at the Autism Genetic Resource Exchange (AGRE) for 612 individuals with a classification of autism and 15 non-spectrum individuals from both AGRE and the Boston Autism Consortium (AC). Our analysis indicated that 8 of the 29 items contained in Module 1 of the ADOS were sufficient to classify autism with 100% accuracy. We further validated the accuracy of this eight-item classifier against complete sets of scores from two independent sources, a collection of 110 individuals with autism from AC and a collection of 336 individuals with autism from the Simons Foundation. In both cases, our classifier performed with nearly 100% sensitivity, correctly classifying all but two of the individuals from these two resources with a diagnosis of autism, and with 94% specificity on a collection of observed and simulated non-spectrum controls. The classifier contained several elements found in the ADOS algorithm, demonstrating high test validity, and also resulted in a quantitative score that measures classification confidence and extremeness of the phenotype. With incidence rates rising, the ability to classify autism effectively and quickly requires careful design of assessment and diagnostic tools. Given the brevity, accuracy and quantitative nature of the classifier, results from this study may prove valuable in the development of mobile tools for preliminary evaluation and clinical prioritization-in particular those focused on assessment of short home videos of children--that speed the pace of initial evaluation and broaden the reach to a significantly larger percentage of the population at risk.

Low molecular weight proteomic information distinguishes metastatic from benign pheochromocytoma.

Metastatic lesions occur in up to 36% of patients with pheochromocytoma. Currently there is no way to reliably detect or predict which patients are at risk for metastatic pheochromocytoma. Thus, the discovery of biomarkers that could distinguish patients with benign disease from those with metastatic disease would be of great clinical value. Using surface-enhanced laser desorption ionization protein chips combined with high-resolution mass spectrometry, we tested the hypothesis that pheochromocytoma pathologic states can be reflected as biomarker information within the low molecular weight (LMW) region of the serum proteome. LMW protein profiles were generated from the serum of 67 pheochromocytoma patients from four institutions and analyzed by two different bioinformatics approaches employing pattern recognition algorithms to determine if the LMW component of the circulatory proteome contains potentially useful discriminatory information. Both approaches were able to identify combinations of LMW molecules which could distinguish all metastatic from all benign pheochromocytomas in a separate blinded validation set. In conclusion, for this study set low molecular mass biomarker information correlated with pheochromocytoma pathologic state using blinded validation. If confirmed in larger validation studies, efforts to identify the underlying diagnostic molecules by sequencing would be warranted. In the future, measurement of these biomarkers could be potentially used to improve the ability to identify patients with metastatic disease.

High-resolution serum proteomic features for ovarian cancer detection.

Serum proteomic pattern diagnostics is an emerging paradigm employing low-resolution mass spectrometry (MS) to generate a set of biomarker classifiers. In the present study, we utilized a well-controlled ovarian cancer serum study set to compare the sensitivity and specificity of serum proteomic diagnostic patterns acquired using a high-resolution versus a low-resolution MS platform. In blinded testing sets, the high-resolution mass spectral data contained multiple diagnostic signatures that were superior to the low-resolution spectra in terms of sensitivity and specificity (P<0.00001) throughout the range of modeling conditions. Four mass spectral feature set patterns acquired from data obtained exclusively with the high-resolution mass spectrometer were 100% specific and sensitive in their diagnosis of serum samples as being acquired from either unaffected patients or those suffering from ovarian cancer. Important to the future of proteomic pattern diagnostics is the ability to recognize inferior spectra statistically, so that those resulting from a specific process error are recognized prior to their potentially incorrect (and damaging) diagnosis. To meet this need, we have developed a series of quality-assurance and in-process control procedures to (a) globally evaluate sources of sample variability, (b) identify outlying mass spectra, and (c) develop quality-control release specifications. From these quality-assurance and control (QA/QC) specifications, we identified 32 mass spectra out of the total 248 that showed statistically significant differences from the norm. Hence, 216 of the initial 248 high-resolution mass spectra were determined to be of high quality and were remodeled by pattern-recognition analysis. Again, we obtained four mass spectral feature set patterns that also exhibited 100% sensitivity and specificity in blinded validation tests (68/68 cancer: including 18/18 stage I, and 43/43 healthy). We conclude that (a) the use of high-resolution MS yields superior classification patterns as compared with those obtained with lower resolution instrumentation; (b) although the process error that we discovered did not have a deleterious impact on the present results obtained from proteomic pattern analysis, the major source of spectral variability emanated from mass spectral acquisition, and not bias at the clinical collection site; (c) this variability can be reduced and monitored through the use of QA/QC statistical procedures; (d) multiple and distinct proteomic patterns, comprising low molecular weight biomarkers, detected by high-resolution MS achieve accuracies surpassing individual biomarkers, warranting validation in a large clinical study.

Mass spectrometry-based proteomics and analyses of serum: a primer for the clinical investigator.

The vocabulary of proteomics and the swiftly-developing, technological nature of the field constitute substantial barriers to clinical investigators. In recent years, mass spectrometry has emerged as the most promising technique in this field. The purpose of this review is to introduce the field of mass spectrometry-based proteomics to clinical investigators, to explain many of the relevant terms, to introduce the equipment employed in this field, and to outline approaches to asking clinical questions using a proteomic approach. Examples of clinical applications of proteomic techniques are provided from the fields of cancer and vasculitis research, with an emphasis on a pattern recognition approach.