Dataset Information

In Silico Labeling: Predicting Fluorescent Labels in Unlabeled Images.

ABSTRACT: Microscopy is a central method in life sciences. Many popular methods, such as antibody labeling, are used to add physical fluorescent labels to specific cellular constituents. However, these approaches have significant drawbacks, including inconsistency; limitations in the number of simultaneous labels because of spectral overlap; and necessary perturbations of the experiment, such as fixing the cells, to generate the measurement. Here, we show that a computational machine-learning approach, which we call "in silico labeling" (ISL), reliably predicts some fluorescent labels from transmitted-light images of unlabeled fixed or live biological samples. ISL predicts a range of labels, such as those for nuclei, cell type (e.g., neural), and cell state (e.g., cell death). Because prediction happens in silico, the method is consistent, is not limited by spectral overlap, and does not disturb the experiment. ISL generates biological measurements that would otherwise be problematic or impossible to acquire.

SUBMITTER: Christiansen EM

PROVIDER: S-EPMC6309178 | biostudies-literature | 2018 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

In Silico Labeling: Predicting Fluorescent Labels in Unlabeled Images.

Christiansen Eric M EM Yang Samuel J SJ Ando D Michael DM Javaherian Ashkan A Skibinski Gaia G Lipnick Scott S Mount Elliot E O'Neil Alison A Shah Kevan K Lee Alicia K AK Goyal Piyush P Fedus William W Poplin Ryan R Esteva Andre A Berndl Marc M Rubin Lee L LL Nelson Philip P Finkbeiner Steven S

Cell 20180412 3

Microscopy is a central method in life sciences. Many popular methods, such as antibody labeling, are used to add physical fluorescent labels to specific cellular constituents. However, these approaches have significant drawbacks, including inconsistency; limitations in the number of simultaneous labels because of spectral overlap; and necessary perturbations of the experiment, such as fixing the cells, to generate the measurement. Here, we show that a computational machine-learning approach, wh ...[more]

PMID: 29656897

Similar Datasets

Project description:The simultaneous measurement of gene expression for thousands of genes in a single analysis by the microarray technology allows researchers to describe transcriptomes in various samples of interest. Problems with variation in data quality derived from microarray experiments are well known and might result from poor RNA quality, background problems, or sub optimal signal strength. To assess variation due to the fluorescent dye chosen, three different dye pairs were tested for labelling of cDNA in gene expression analysis experiments on a porcine immune focused oligonucleotide microarray (POM3). This in-house oligonucleotide microarray allowed a direct comparison of background fluorescence, Median signal intensities, numbers of spots detected, and resistance to photobleaching between different dye pairs. We tested Alexa Flour 546/647, Cy3/Cy5 as well as Oyster 550/650 all from Genisphere Inc., Hatfield, PA, USA. Keywords: Comparison of fluorescent dyes Each dye pairs were hybridized on two slides. The same two samples were compared on all the slides used in the present study. The liver sample from a healthy animal was labeled with the high wavelength dyes (Oyster 650/Alexa 647/Cy5) and the liver sample from the sick animal was labeled with the low wavelength dyes (Oyster 550/Alexa 546/Cy3). Each slide was scanned 3 times, immediately after hybridization (first round), after 1 month of storage in darkness and after 24 hours on the bench (12 hours of daylight and 12 hours of artificial light). Total RNA from the infected and control liver samples were extracted using RNeasy midi kit (Qiagen, Denmark) and DNase treated using RNase-Free DNase Set (Qiagen). 3DNATM Array 900 expression array detection kits (Genisphere Inc.) were used for the labeling and cDNA synthesis reaction of the RNA in the present study. Three different pairs of fluorescent dyes Oyster 550/650 (Genisphere Inc.), Alexa 546/647 (Genisphere Inc.), and Cy3/Cy5 (Genisphere Inc.) were used in separate labeling reactions. Labeling was done according to the manufacturers protocol for large-scale cDNA synthesis. For the cDNA synthesis 9.2Âµg total RNA from each liver sample was used and mixed with 20U AmpliQ RTenzyme and 10x first strand buffer (Bie og Berntsen, RÃ¸dovre, Denmark). Hybridization and washing were performed according to the manufacturers instructions (Genisphere) using Corning hybridization chambers. For one cDNA hybridization reaction; 6.5Âµl cDNA from each synthesis, 0.5Âµl Salmon sperm (10Âµg/Âµl) and 13.5Âµl 2 x Formamide-Based Hybridization Buffer (3DNA Array 900, Genisphere) was used. A total hybridization mix of 29Âµl was applied under a 22I x 25 mm LifterSlip (Erie Scientific, Portsmouth, NH, USA) carefully avoiding air bubbles. Slides were incubated at 44oC in a water bath over night. Wash buffer 1 (3DNA Array 900, Genisphere) was preheated to 44oC for the post cDNA hybridization wash. For two 3DNA hybridization reaction mixes; 2.5Âµl of each Capture reagent and 26Âµl SDS-Based hybridization buffer was mixed to a final volume of 52Âµl. 26Âµl 3DNA hybridization mix was applied to each slide, and incubated in darkness in a water bath at 50oC for 4 hours. Wash buffer 1 was preheated to 60oC for the post 3DNA hybridization wash. Slides were scanned on a CCD ArrayWoRxe auto (Applied Precision, Issaquah, WA, USA) using different exposure times (0.3 for 595 nm and 1.2 for 685 nm).

Dataset Information

In Silico Labeling: Predicting Fluorescent Labels in Unlabeled Images.

Publications

In Silico Labeling: Predicting Fluorescent Labels in Unlabeled Images.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets