Dataset of Two-Dimensional Gel Electrophoresis Images of Acute Myeloid Leukemia Patients before and after Induction Therapy

Abstract

Acute myeloid leukemia (AML) is a malignant disorder of the hematopoietic stem and progenitor cells, which results in the build-up of immature blasts in the bone marrow and eventually in the peripheral blood of affected patients. Accurately assessing a patient´s prognosis is very important for clinical management of the disease, which is why there are several prognostic factors such as age, performance status at diagnosis, platelet count, serum creatinine and albumin that are taken into account by the clinician when deciding the course of treatment. However, proteomic changes related to treatment response in this patient group have not been widely explored. Here, we make available a set of 22 two-dimensional gel electrophoresis (2DGE) images obtained from the peripheral blood samples of 11 patients with AML, taken at the time of diagnosis and after induction therapy (approximately 21–28 days after starting treatment). The same set of 2DGE images is also made available after a preprocessing stage (an additional 22 2DGE pre-processed images), which was performed using algorithms developed in Python, in order to improve the visualization of characteristic spots and facilitate proteomic analysis of this type of images.

Dataset: The dataset will be published as a supplement to this paper, so this field will be filled by the editors of the journal.

Dataset License: CC-BY 4.0

1. Summary

According to the Global Cancer Observatory (Globocan 2018), each year, 437,033 patients worldwide are diagnosed with some type of leukemia, and 309,006 people die from this disease. Acute myeloid leukemia (AML) is a type of leukemia that mainly occurs in older adults; 42% of Americans diagnosed with AML are over 65 years of age, and their diagnosis is rarely made before 40 years of age, although cases have progressively increased over time [1]. AML is the result of an accumulation of acquired genetic alterations in the DNA of hematopoietic progenitor cells, and accurately assessing a patient’s prognosis is very important for clinical management of the disease. The patient’s cytogenetic profile is currently the strongest prognostic factor. For example, a complex karyotype, monosomy 5 or 7, t(6;9), inv(3), or 11q changes, other than t(9;11), have all been associated with a significantly lower response to treatment and overall survival [2]. It is clear that genetic studies are very valuable; however, isolated from a context in which thousands of proteins mediate cellular function, this prognostic model is not complete.

The images of this dataset were obtained by two-dimensional gel electrophoresis (2DGE), a technique that separates proteins according to their isoelectric point and molecular weight [3], followed by protein staining and image capture. Often, 2DGE images include anomalies [4,5] such as vertical lines, horizontal lines, diffuse points, and noise, among others, which make it difficult to identify spots that contain valuable information. Therefore, a preprocessing stage is often necessary in order to discriminate stains and noise from real protein spots [6]. Omitting this stage can affect the interpretation of the data, as noise could be identified as false protein spots [7]. Image preprocessing is responsible for reducing or correcting these irregularities in 2DGE images. The authors have implemented an approach that integrates the techniques of image normalization, noise reduction by non-linear techniques, and background correction [4,8], sequentially applying the following structure: adaptive piecewise histogram equalization for image normalization, a geometric nonlinear diffusion filter (GNDF) for filtering, and multilevel thresholding for background correction, obtaining favorable results [9].

2. Data Description

The database consists of a set of 22 2DGE images obtained from the peripheral blood samples of 11 patients with acute myeloid leukemia. Of these, 11 images correspond to samples taken at the time of diagnosis, and the other 11 correspond to samples taken from the same patients after induction therapy (approximately 21–28 days after starting treatment). Images named with the suffix BEFORE refer to 2DGE images of samples taken at the time of diagnosis (before treatment), while images named with the suffix AFTER correspond to 2DGE images of samples taken after treatment. These 22 images are also made available with the preprocessing stage applied, to which the prefix PREPROC has been applied. Each image in the database is in tagged image file format (TIFF) format with a resolution of 300 dots per inch (DPI). In total, the database, which can be found in the Supplementary Materials, contains 44 images (22 raw 2DGE images and 22 pre-processed 2DGE images). The characteristics corresponding to each image are summarized in

Table 1. Image data description.

2DGE Image	2DGE Image Size	2DGE Pre-Processed Image	2DGE Pre-Processed Image Size	Width (Pixels)	Height (Pixels)
HMUA02_BEFORE	564 KB	PREPROC_HMUA02_BEFORE	1.63 MB	965	757
HMUA02_AFTER	530 KB	PREPROC_HMUA02_AFTER	1.54 MB	1006	803
HMUA03_BEFORE	607 KB	PREPROC_HMUA03_BEFORE	1.94 MB	972	820
HMUA03_AFTER	554 KB	PREPROC_HMUA03_AFTER	1.73 MB	974	798
HMUA04_BEFORE	589 KB	PREPROC_HMUA04_BEFORE	1.65 MB	1021	842
HMUA04_AFTER	607 KB	PREPROC_HMUA04_AFTER	1.88 MB	1035	866
HMUA05_BEFORE	556 KB	PREPROC_HMUA05_BEFORE	1.90 MB	1018	810
HMUA05_AFTER	558 KB	PREPROC_HMUA05_AFTER	1.65 MB	1012	788
HMUA010_BEFORE	584 KB	PREPROC_HMUA010_BEFORE	2.21 MB	1021	838
HMUA010_AFTER	585 KB	PREPROC_HMUA010_AFTER	2.01 MB	1012	828
HMUA011_BEFORE	527 KB	PREPROC_HMUA011_BEFORE	1.70 MB	985	777
HMUA011_AFTER	506 KB	PREPROC_HMUA011_AFTER	1.78 MB	974	781
HMUA012_BEFORE	603 KB	PREPROC_HMUA012_BEFORE	2.14 MB	971	798
HMUA012_AFTER	430 KB	PREPROC_HMUA012_AFTER	1.75 MB	969	775
HMUA013_BEFORE	498 KB	PREPROC_HMUA013_BEFORE	1.73 MB	1024	803
HMUA013_AFTER	561 KB	PREPROC_HMUA013_AFTER	1.85 MB	1012	820
HMUA015_BEFORE	552 KB	PREPROC_HMUA015_BEFORE	1.67 MB	1054	857
HMUA015_AFTER	576 KB	PREPROC_HMUA015_AFTER	1.88 MB	1046	870
HMUA017_BEFORE	501 KB	PREPROC_HMUA017_BEFORE	1.91 MB	983	757
HMUA017_AFTER	476 KB	PREPROC_HMUA017_AFTER	1.84 MB	1102	921
HMUA018_BEFORE	432 KB	PREPROC_HMUA018_BEFORE	1.38 MB	1111	858
HMUA018_AFTER	526 KB	PREPROC_HMUA018_AFTER	1.79 MB	1036	795

.

3. Methods

3.1. Patients

Peripheral blood was obtained from 11 newly diagnosed patients with de novo AML at Hospital Manuel Uribe Angel in Colombia. Two blood samples were taken from each patient: at the time of diagnosis (before the start of chemotherapy) and once again after completion of the first round of induction therapy, which was typically 2–3 weeks after induction or when neutrophil and platelet recovery was achieved. Relevant clinical information of the patients involved in this study is summarized in

Table 2. Clinical Information.

Patient	Age	Sex	Karyotype	AML Subtype 1	Blasts (i) 2	Induction Protocol	Blasts (f) 3	Response to Induction
HMUA_02	35	M	Normal	M3	86%	PETHEMA	1%	CR 4
HMUA_03	56	M	Normal	M4	65%	7 + 3	4.5%	CR
HMUA_04	67	M	Normal	M4	14%	FLUGA	35%	Resistant
HMUA_05	42	F	T(8:21)	M3	85%	PETHEMA	1.2%	CR
HMUA_10	18	F	Missing	M1/M2	74%	7 + 3	0.5%	CR
HMUA_11	65	F	CCR 5	M4	42%	7 + 3	11%	PR 6
HMUA_12	53	F	T(8:21)	M2	20%	7+3	1.3%	CR
HMUA_13	46	M	Normal	M2	68%	7+3	35.4%	Resistant 7
HMUA_15	50	M	CCR	M1/M2	47%	7+3	8.4%	PR
HMUA_17	67	M	T(8:21)	M4	66%	7+3	0.27%	CR
HMUA_18	18	F	Normal	M1/M2	28%	7+3	0.6%	CR

¹ According to the French–American–British (FAB) classification system. ² (i): initial blast count, before induction therapy. ³ (f): final blast count, after induction therapy. ⁴ CR: complete remission, defined as <5% blasts in bone marrow after induction therapy. ⁵ CCR: complex chromosome rearrangement. ⁶ PR: partial response, defined as 5–20% blasts in bone marrow after induction therapy. ⁷ Resistance to therapy, defined by >20% blasts in bone marrow after induction therapy.

.

3.2. Protein Extraction

Peripheral blood mononuclear cells (PBMCs) were isolated from the blood samples by standard density gradient centrifugation with a Ficoll Histopaque^®-1077 (Sigma-Aldrich, St. Louis, MO, USA). In order to extract proteins from the PBMCs, the cells were lysed (0.5% Triton x-100, 50 mM Tris-HCL pH 8.0, 150 mM NaCL, 1 mM Ethylenediaminetetraacetic acid (EDTA), protease inhibitors) and the proteins precipitated in a 20% (v/v) final concentration of trichloroacetic acid. The protein pellet was resuspended in a rehydration buffer (7 M urea, 2% 3-cholamidopropyl dimethylammonio 1-propanesulfonate (CHAPS), 0.5% carrier ampholytes) and stored at −70 °C.

3.3. Two-Dimensional Gel Electrophoresis

Proteins (50 µg) were loaded by passive rehydration onto 7 cm ZOOM^® immobilized pH gradient (IPG) strips with a pH of 3–10 NL (ThermoFisher Scientific, Waltham, MA, USA) at room temperature. Isoelectric focusing was carried out using the following voltage ramp: 100 V for 1 h, 150 V for 1 h, 200 V for 5 min, 450 V for 5 min, 600 V for 5 min, 750 V for 5 min, 950 V for 5 min, 1200 V for 10 min, 1400 V for 10 min, 1600 V for 10 min, and 2000 V for 45 min. The IPG strips were then reduced with 100 mM Dithiothreitol (DTT) and alkylated with 2.5% iodacetamide, according to the manufacturer’s recommended protocol. After this, the IPG strips were loaded onto SDS-PAGE NuPAGE^™ Novex^™ 4–12% Bis-Tris protein gels 1.5 mm in size (ThermoFisher Scientific) and run at 200 V for 45 min. After electrophoresis, these were stained with SYPRO^® Ruby (Invitrogen^™, ThermoFisher Scientific), and the gel images were acquired using the ChemiDoc^™ MP System (Biorad).

3.4. Image Pre-Processing

This step was performed in order to mitigate anomalies due to the acquisition routines and improve spot detection. The approach proposed in [9] was applied, integrating the following techniques for image normalization, noise reduction, and background correction: adaptive piecewise histogram equalization, a geometric nonlinear diffusion filter (GNDF), and multilevel thresholding. The algorithm was executed in Python, which is an open-source programming language, with free access to permanent online support through a considerable number of available libraries, accelerating the creation of multi-stage structure codes with the aim of obtaining consistent, reliable, and potentially integrable results.